Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+lfm2_5_350m_mlx_4w.pte filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,183 @@
+---
+license: cc-by-nc-4.0
+license_link: https://huggingface.co/LiquidAI/LFM2.5-350M  # confirm and adjust to base-model license
+library_name: executorch
+base_model: LiquidAI/LFM2.5-350M
+pipeline_tag: text-generation
+tags:
+  - ExecuTorch
+  - MLX
+  - apple-silicon
+  - text-generation
+  - lfm2.5
+  - formatter
+  - dictation
+  - on-device
+datasets:
+  - edinburghcstr/ami
+---
+
+# LFM2.5-350M — ExecuWhisper Formatter
+
+A fine-tuned [LiquidAI/LFM2.5-350M](https://huggingface.co/LiquidAI/LFM2.5-350M) tuned to clean up spoken dictation: remove disfluencies, restore casing and punctuation, infer light list structure, and **refuse to "answer"** the dictation even when the user's voice ends in a question.
+
+Built for the [ExecuWhisper](https://github.com/meta-pytorch/executorch-examples/tree/main/execuwhisper/macos) macOS dictation app. Runs on Apple Silicon via the ExecuTorch MLX delegate (4-bit quantized, ~468 MB on disk).
+
+## What this is (and isn't)
+
+This model is a **dictation cleaner**, not a chat assistant. Given:
+> _"um does it feel like real time processing"_
+
+It outputs:
+> _"Does it feel like real-time processing?"_
+
+It will *not* answer the question, *not* add information, and *not* summarize. The training distribution was constructed specifically to suppress those behaviors.
+
+It also won't help you with arbitrary text-generation tasks. For those, use the [base model](https://huggingface.co/LiquidAI/LFM2.5-350M) directly.
+
+## Files
+
+| Path | Size | Purpose |
+|---|---:|---|
+| `lfm2_5_350m_ft.pt` | 1.42 GB | fp32 fine-tuned checkpoint (re-quantize / re-train baseline) |
+| `lfm2_5_350m_mlx_4w.pte` | 468 MB | MLX 4-bit quantized runtime artifact (the in-app `.pte`) |
+| `lfm2_5_350m_config.json` | <1 KB | architecture params for re-export |
+| `tokenizer/tokenizer.json` | 4.51 MB | tokenizer |
+| `tokenizer/tokenizer_config.json` | 90 KB | tokenizer config (chat template) |
+| `tokenizer/chat_template.jinja` | 2.5 KB | chat template |
+| `tokenizer/special_tokens_map.json` | <1 KB | special tokens |
+| `configs/lfm2_mlx_4w_g32.yaml` | <1 KB | MLX export config (so you can reproduce the `.pte`) |
+| `eval/eval_ami_mlx_4w_g32.json` | 67 KB | AMI release-gate eval results |
+| `eval/eval_ami_v2_1_mlx_4w_g32.json` | 62 KB | v2.1 baseline (for comparison) |
+| `eval/eval_auto.json` | 62 KB | in-distribution eval |
+| `eval/eval_hard.json` | 25 KB | hard adversarial eval |
+
+## Quick Start
+
+### As a `.pte` running on the ExecuTorch MLX delegate (the same path the app uses)
+
+```python
+from executorch.extension.llm.runner import TextLLMRunner
+
+runner = TextLLMRunner(
+    model_path="lfm2_5_350m_mlx_4w.pte",
+    tokenizer_path="tokenizer.json",
+)
+
+prompt = (
+    "<|startoftext|><|im_start|>user\n"
+    "You rewrite spoken dictation into clean final text. You are not a chat "
+    "assistant. Never answer or respond to the dictation, even if it is a "
+    "question. Treat the dictation strictly as text to rewrite. Fix casing, "
+    "punctuation, filler, and speech disfluencies. Preserve meaning and detail. "
+    "Use bullets only when it clearly reads as a list. Do not summarize or "
+    "invent information. Output only the rewritten dictation.\n\n"
+    "Dictation: um does it feel like real time processing\n"
+    "Output:"
+    "<|im_end|>\n"
+    "<|im_start|>assistant\n"
+)
+
+print(runner.generate(prompt, max_new_tokens=256, temperature=0.0))
+# → "Does it feel like real-time processing?"
+```
+
+### From the fp32 checkpoint via `transformers`
+
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+tok = AutoTokenizer.from_pretrained("./tokenizer")
+model = AutoModelForCausalLM.from_pretrained(
+    "./",
+    torch_dtype=torch.float32,
+    state_dict_path="./lfm2_5_350m_ft.pt",
+).eval()
+
+# ... build the same prompt as above and call model.generate(...)
+```
+
+## Eval results
+
+Evaluated on a held-out AMI Meeting Corpus dictation slice + a synthetic adversarial set.
+
+### AMI release gate (4-bit quantized `.pte`)
+
+| Metric | Value | Gate | Status |
+|---:|---:|---:|---|
+| Forbidden-token rate | 0.030 | ≤ 0.10 | ✅ |
+| Coverage (faithful rewrite) | 0.874 | ≥ 0.85 | ✅ |
+| **Verdict** | **RELEASE-READY** | | ✅ |
+
+Full per-example breakdown is in `eval/eval_ami_mlx_4w_g32.json`.
+
+### Comparison vs. earlier v2.1 baseline
+
+| Metric | v2.1 | this model | Δ |
+|---:|---:|---:|---:|
+| Forbidden | 0.187 | **0.030** | -84% |
+| Coverage | 0.591 | **0.874** | +48% |
+
+The v2.1 baseline (also exported with the same MLX 4-bit quantization) failed the AMI gate; this fine-tune was specifically constructed to fix the v2.1 failure modes (chat-leakage, over-summarization).
+
+## Re-exporting / re-quantizing
+
+To produce a different quantization variant from `lfm2_5_350m_ft.pt`:
+
+1. Check out the LFM2.5 MLX export pipeline: [`pytorch/executorch#19195`](https://github.com/pytorch/executorch/pull/19195).
+2. Use `configs/lfm2_mlx_4w_g32.yaml` as a starting point.
+3. Run the LFM2.5 export Makefile target with your edited config:
+   ```bash
+   cd ~/executorch
+   make lfm_2_5-mlx LFM_CONFIG=path/to/your_config.yaml LFM_CHECKPOINT=path/to/lfm2_5_350m_ft.pt
+   ```
+
+## Fine-tuning your own
+
+To adapt this model to a new domain (medical, legal, multilingual dictation), follow the **[Unsloth LFM2.5 fine-tuning tutorial](https://unsloth.ai/docs/models/tutorials/lfm2.5)**. The tutorial covers SFT + LoRA, hyperparameter selection, and export.
+
+The training data for this model was a mix of:
+
+- ~1,350 synthetic dictation pairs (clean target → noisified spoken input via filler/disfluency injection + casing distortion).
+- ~706 dictation-style turns extracted from the AMI Meeting Corpus.
+
+The synthetic pipeline and AMI extraction code are not yet open-sourced; the eval splits in `eval/` are the publicly verifiable artifacts.
+
+## Limitations
+
+- **Self-corrections** — over-summarizes "actually no — make it tomorrow" patterns; sometimes drops the corrected clause.
+- **Email sign-offs** — occasionally drops the closing name in template-style sign-offs ("Best, Younghan" → "Best,").
+- **Long context** — the in-app pipeline chunks transcripts longer than ~30 words. Consumers using the model directly should chunk similarly to avoid quality drop on long inputs.
+- **English only** — trained on English dictation; behavior on other languages is undefined.
+- **Not a chat model** — will refuse / ignore questions, by design.
+
+## License & acknowledgements
+
+This derivative inherits the [LiquidAI/LFM2.5-350M](https://huggingface.co/LiquidAI/LFM2.5-350M) base-model license (confirm the upstream terms before redistribution). Eval data is derived from the [AMI Meeting Corpus](https://groups.inf.ed.ac.uk/ami/corpus/) (CC-BY-4.0).
+
+Thanks to:
+
+- **LiquidAI** — for releasing LFM2.5-350M and the LFM architecture.
+- **Apple MLX team** — for `mlx` and the MLX delegate inside ExecuTorch.
+- **PyTorch / ExecuTorch team** — for the runtime and the export pipeline.
+- **University of Edinburgh** and the AMI corpus contributors — for the dictation eval source.
+- **Unsloth** — for the fine-tuning recipe.
+
+## Citation
+
+```bibtex
+@software{execuwhisper_formatter2026,
+  title = {LFM2.5-350M ExecuWhisper Formatter},
+  author = {Han, Younghan},
+  year = {2026},
+  url = {https://huggingface.co/younghan-meta/LFM2.5-350M-ExecuWhisper-Formatter},
+  note = {Fine-tuned LFM2.5-350M dictation cleaner; 4-bit MLX quantization for Apple Silicon}
+}
+```
+
+## Companion projects
+
+- **[ExecuWhisper macOS app](https://github.com/meta-pytorch/executorch-examples/tree/main/execuwhisper/macos)** — the consuming dictation app.
+- **[pytorch/executorch](https://github.com/pytorch/executorch)** — runtime, MLX delegate, export pipeline.
+- **[Unsloth LFM2.5 tutorial](https://unsloth.ai/docs/models/tutorials/lfm2.5)** — recommended fine-tuning path.

configs/lfm2_mlx_4w_g32.yaml

@@ -0,0 +1,16 @@
+base:
+  metadata: '{"get_bos_id": 1, "get_eos_ids": [7]}'
+model:
+  use_kv_cache: True
+  use_sdpa_with_kv_cache: True
+  dtype_override: fp32
+quantization:
+  qmode: "4w"
+  group_size: 32
+  use_hqq: True
+export:
+  max_seq_length: 2048
+  max_context_length: 2048
+backend:
+  mlx:
+    enabled: True

eval/eval_ami_mlx_4w_g32.json

@@ -0,0 +1,1151 @@
+{
+  "overall": {
+    "n": 100,
+    "forbidden_rate": 0.03,
+    "exact_match_rate": 0.08,
+    "mean_similarity": 0.9152703164555943,
+    "target_coverage": 0.8740067740060208,
+    "gates": {
+      "forbidden": true,
+      "coverage": true
+    },
+    "mean_tokens_per_sec": 533.62,
+    "verdict": "RELEASE-READY"
+  },
+  "by_category": {
+    "prose": {
+      "n": 83,
+      "forbidden_rate": 0.0241,
+      "exact_match_rate": 0.0843,
+      "mean_similarity": 0.9153853637294971,
+      "target_coverage": 0.8792753407312947
+    },
+    "meeting": {
+      "n": 14,
+      "forbidden_rate": 0.0714,
+      "exact_match_rate": 0.0714,
+      "mean_similarity": 0.905432248000447,
+      "target_coverage": 0.8518207435355283
+    },
+    "intent": {
+      "n": 3,
+      "forbidden_rate": 0.0,
+      "exact_match_rate": 0.0,
+      "mean_similarity": 0.9579983280016382,
+      "target_coverage": 0.8317779034690799
+    }
+  },
+  "forbidden_hits": {
+    "here's the": 1,
+    "here is": 1,
+    "sure,": 1
+  },
+  "rows": [
+    {
+      "i": 0,
+      "category": "prose",
+      "transcript": "Oh, I think it is uh um s starting with um y European, I think.",
+      "target": "Oh, I think it's starting with European, I think.",
+      "output": "Oh, I think it is starting with YELEAN, I think.",
+      "exact": false,
+      "similarity": 0.845,
+      "target_coverage": 0.556,
+      "forbidden": []
+    },
+    {
+      "i": 1,
+      "category": "meeting",
+      "transcript": "You could just like sort of zoom back out till you've got the full text for that segment or that meeting rather than a summary.",
+      "target": "You could just, like, sort of zoom back out so you got the full text for that segment or that meeting rather than a summary.",
+      "output": "You could just, like, sort of, or, sort of, back out till you've got the full text for that segment or that meeting rather than a summary.",
+      "exact": false,
+      "similarity": 0.916,
+      "target_coverage": 0.792,
+      "forbidden": []
+    },
+    {
+      "i": 2,
+      "category": "prose",
+      "transcript": "Um so uh I guess just to confirm that we know what we're doing in the next well, in the thirty minutes after lunch anyway.",
+      "target": "So just to confirm that we know what we're doing in the\u2014well, in the thirty minutes after lunch anyway.",
+      "output": "So I guess just to confirm that we know what we're doing in the next\u2014well, in the thirty minutes after lunch, anyway.",
+      "exact": false,
+      "similarity": 0.936,
+      "target_coverage": 0.9,
+      "forbidden": []
+    },
+    {
+      "i": 3,
+      "category": "intent",
+      "transcript": "And you might want to talk to Alex to find out if somebody's already done this turn redwood stuff into tree bank stuff so that they can compare",
+      "target": "And you might wanna talk to Alex to find out if somebody's already done this, turn Redwoods stuff into Treebank stuff, so that they can compare.",
+      "output": "And you might want to talk to Alex to find out if somebody's already done this turn redwood stuff into tree bank stuff, so that they can compare.",
+      "exact": false,
+      "similarity": 0.955,
+      "target_coverage": 0.769,
+      "forbidden": []
+    },
+    {
+      "i": 4,
+      "category": "prose",
+      "transcript": "Because it's quite easy in Pearl as well. It's just a line of code for counting all the words and Yeah, it's it's by hashes.",
+      "target": "Because it's quite easy in Perl as well, it's just a line of code for counting all the words, and yeah, it's by hashes.",
+      "output": "Because it's quite easy in Pearl as well, it's just a line of code for counting all the words, and yeah, it's by hashes.",
+      "exact": false,
+      "similarity": 0.996,
+      "target_coverage": 0.923,
+      "forbidden": []
+    },
+    {
+      "i": 5,
+      "category": "prose",
+      "transcript": "If if your intention is to move the red triangle to sit on top of the thing already constructed, you would tend to look back and forth.",
+      "target": "If your intention is to move the red triangle to sit on top of the thing already constructed, you would tend to look back and forth.",
+      "output": "If your intention is to move the red triangle to sit on top of the thing already constructed, you would tend to look back and forth.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 6,
+      "category": "prose",
+      "transcript": "Um one thing that would be convenient for all sorts of things and which I haven't found out about it is how to do tool tips in an arbitrary way in in in swing graphic displays.",
+      "target": "One thing that would be convenient for all sorts of things, and which I haven't found out about, is how to do tooltips in an arbitrary way in Swing graphic displays.",
+      "output": "One thing that would be convenient for all sorts of things, and which I haven't found out about, is how to do tool tips in an arbitrary way in in-swing graphic displays.",
+      "exact": false,
+      "similarity": 0.982,
+      "target_coverage": 0.935,
+      "forbidden": []
+    },
+    {
+      "i": 7,
+      "category": "prose",
+      "transcript": "What we've finally just uh at twelve o'clock today, looks like we've got it sussed, um, is to get one channel one microphone being the left channel and one being the right channel to merge into a stereo single file.",
+      "target": "What we've finally just at twelve o'clock today, looks like we've got it sussed, is to get one channel, one microphone being the left channel, and one being the right channel, to merge into a stereo single file.",
+      "output": "What we've finally just at twelve o'clock today, looks like we've got it\u2014sussed, is to get one channel one microphone, be the left channel and one being the right channel to merge into a stereo single file.",
+      "exact": false,
+      "similarity": 0.978,
+      "target_coverage": 0.975,
+      "forbidden": []
+    },
+    {
+      "i": 8,
+      "category": "prose",
+      "transcript": "I'm just thinking of you know, ill utterance consisting entirely of laughter and then putting like punctuation in like ha ha ha ha just",
+      "target": "I'm just thinking of, you know, an utterance consisting entirely of laughter, and then putting like punctuation in, like...",
+      "output": "I'm just thinking of you, know, ill utterance consisting entirely of laughter, then putting, like, punctuation in, like, ha-ha-ha, just...",
+      "exact": false,
+      "similarity": 0.889,
+      "target_coverage": 0.789,
+      "forbidden": []
+    },
+    {
+      "i": 9,
+      "category": "prose",
+      "transcript": "So we could zip we could zip through this and look at all the um all the interactions in which people built subcontracts first, which we expect them to do as a wild hypothesis more when they have verbal communication than when they don't.",
+      "target": "So we could zip through this and look at all the interactions in which people build subconstructs first, which we expect them to do, but as a wild hypothesis more when they have verbal communication than when they don't.",
+      "output": "So we could zip through this and look at all the interactions in which people built subcontracts first, which we expect them to do as a wild hypothesis, more when they have verbal communication than when they don't.",
+      "exact": false,
+      "similarity": 0.933,
+      "target_coverage": 0.872,
+      "forbidden": []
+    },
+    {
+      "i": 10,
+      "category": "prose",
+      "transcript": "I'm I I'm getting quite lost um at the moment because um what's um our difference between the um s um uh the importance measure and the skimming?",
+      "target": "I'm getting quite lost at the moment because what's our difference between the importance measure and the skimming?",
+      "output": "I'm getting quite lost at the moment because what's our difference between the importance measure and the skimming.",
+      "exact": false,
+      "similarity": 0.991,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 11,
+      "category": "prose",
+      "transcript": "Which if you just said that it's gone Peterson absorption from neutral hydrogen rather than i if it was a if it was hydrogen absorption",
+      "target": "Which, if you just said that it's Gunn-Petersen absorption from neutral hydrogen rather than\u2014if it was a\u2014if it was hydrogen absorption.",
+      "output": "Which, if you just said that it's gone, Peterson absorption from neutral hydrogen rather than if it was hydrogen absorption?",
+      "exact": false,
+      "similarity": 0.903,
+      "target_coverage": 0.792,
+      "forbidden": []
+    },
+    {
+      "i": 12,
+      "category": "prose",
+      "transcript": "So what I'm having is so we we had a we had a long problem about not finding out why the time syncing up between the relative time in a in a sort of in a in a summarized section and the absolute time didn't correspond and stuff.",
+      "target": "So what I'm having is, so we had a long problem about not finding out why the time-syncing up between the relative time in a sort of summarised section and the absolute time didn't correspond and stuff.",
+      "output": "So what I'm having is so we had a long problem about not finding out why the time syncing up between the relative time in a sort of in a summary section and the absolute time didn't correspond, and stuff.",
+      "exact": false,
+      "similarity": 0.936,
+      "target_coverage": 0.949,
+      "forbidden": []
+    },
+    {
+      "i": 13,
+      "category": "meeting",
+      "transcript": "So interactivity is a concept is a concept that I'm using to distinguish the two uh uh extremes of uh of interact uh of discourse in meetings.",
+      "target": "So 'interactivity' is a concept that I'm using to distinguish the two extremes of discourse in meetings.",
+      "output": "Interactivity is a concept that I'm using to distinguish the two extremes of interactivity in meetings.",
+      "exact": false,
+      "similarity": 0.879,
+      "target_coverage": 0.824,
+      "forbidden": []
+    },
+    {
+      "i": 14,
+      "category": "intent",
+      "transcript": "And um so far um I think uh it's uh I think it it's uh still um about the s about the same level of segmentation we are talking here.",
+      "target": "And so far I think it's still about the same level of segmentation we're talking here.",
+      "output": "And so far, I think, it's still about the same level of segmentation, we are talking here.",
+      "exact": false,
+      "similarity": 0.966,
+      "target_coverage": 0.882,
+      "forbidden": []
+    },
+    {
+      "i": 15,
+      "category": "prose",
+      "transcript": "As long as other people don't for some reason, I don't know, disconnect some bits of the equipment, there's no reason why they should.",
+      "target": "That as long as other people don't, for some reason\u2014I don't know, disconnect some bits of the equipment\u2014there's no reason why they should.",
+      "output": "As long as other people don't for some reason, I don't know, disconnect some bits of the equipment, there's no reason why they should.",
+      "exact": false,
+      "similarity": 0.949,
+      "target_coverage": 0.963,
+      "forbidden": []
+    },
+    {
+      "i": 16,
+      "category": "prose",
+      "transcript": "It's fairly general. Um and I guess the purpose of this is also not to restrict you in in the way you're thinking, like um, voice recognition.",
+      "target": "It's fairly general, and I guess the purpose of this is also not to restrict you in the way you're thinking, like voice recognition.",
+      "output": "It's fairly general, and I guess the purpose of this is also not to restrict you in the way you're thinking, like, voice recognition.",
+      "exact": false,
+      "similarity": 0.996,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 17,
+      "category": "prose",
+      "transcript": "W within the sentence that you're sort of currently looking at, um feel free to correct it that way that you like it's not only about resolution but sort of like coming up with a corrected version on these on these places.",
+      "target": "Within the sentence that you sort of currently looking at, feel free to correct it that way, that you'll do, you know, like, it's not only about resolution but it's sort of like coming up with a corrected version based on these.",
+      "output": "W, within the sentence that you're sort of currently looking at, feel free to correct it that way that you like it's not only about resolution but sort of, like, coming up with a corrected version on these places.",
+      "exact": false,
+      "similarity": 0.109,
+      "target_coverage": 0.773,
+      "forbidden": []
+    },
+    {
+      "i": 18,
+      "category": "prose",
+      "transcript": "And there will also be presumably some kind of task hierarchy modeling going on in I mean,'cause the whole browser working group um final um sort of presentation on that second day was about",
+      "target": "And there will also be, presumably, some kind of task hierarchy modelling going on, in\u2014I mean, 'cause the whole browser working group final, sort of, presentation on that second day was about...",
+      "output": "And there will also be presumably some kind of task hierarchy modelling going on, in\u2014I mean, 'cause the whole browser working group final sort of presentation on that second day was about.",
+      "exact": false,
+      "similarity": 0.984,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 19,
+      "category": "prose",
+      "transcript": "Uh right, well basically um high priority for any animal for me is that they be willing to take a lot of physical affection from their family.",
+      "target": "Right, well, basically high priority for any animal, for me, is that they be willing to take a lot of physical affection from their family.",
+      "output": "Well, basically, high priority for any animal, for me, is that they be willing to take a lot of physical affection from their family.",
+      "exact": false,
+      "similarity": 0.963,
+      "target_coverage": 0.958,
+      "forbidden": []
+    },
+    {
+      "i": 20,
+      "category": "prose",
+      "transcript": "It took like the third time till I thought, Oh, you're saying F five.",
+      "target": "It took, like, the third time till I thought, oh, you're saying F-five.",
+      "output": "It took, like, the third time, till I thought, you're saying five.",
+      "exact": false,
+      "similarity": 0.949,
+      "target_coverage": 0.714,
+      "forbidden": []
+    },
+    {
+      "i": 21,
+      "category": "prose",
+      "transcript": "Because, you know, algorithmically you're you're already putting together I mean this is essentially already, you know, these fixations with stuff in between.",
+      "target": "Because, you know, algorithmically, you're already putting together\u2014I mean, this is essentially already, you know, these fixations with stuff in between.",
+      "output": "Because, you know, algorithmically you're already putting together\u2014I mean, this is essentially already, you know, these fixations with stuff in between.",
+      "exact": false,
+      "similarity": 0.997,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 22,
+      "category": "prose",
+      "transcript": "Yeah. Uh right, damp life enough. See again, that's uh these are source types. Um I mean it's not a source type.",
+      "target": "Right, damped Lyman\u2014see, again, that's\u2014these are source types\u2014the, I mean, it's not a source type.",
+      "output": "Yeah, right, damp life enough, see again, that's these are source types, I mean, it's not a source type.",
+      "exact": false,
+      "similarity": 0.832,
+      "target_coverage": 0.737,
+      "forbidden": []
+    },
+    {
+      "i": 23,
+      "category": "prose",
+      "transcript": "So mm so for dialogue, yeah, mm there would be uh some kind of uh interaction structure.",
+      "target": "So for dialogue, yeah, there would be some kind of interaction structure.",
+      "output": "So, for dialogue, yeah, there would be some kind of interaction structure.",
+      "exact": false,
+      "similarity": 0.993,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 24,
+      "category": "prose",
+      "transcript": "Well, I could look'cause I I've got that there's the list, isn't there, of postgraduates that I maintain in in linguistics.",
+      "target": "Well, I could look, 'cause I've got\u2014that's the list, isn't there, of postgraduates I maintain in linguistics.",
+      "output": "Well, I could look, 'cause I've got that\u2014there's the list, isn't there, of postgraduates that I maintain in linguistics.",
+      "exact": false,
+      "similarity": 0.943,
+      "target_coverage": 0.9,
+      "forbidden": [
+        "here's the"
+      ]
+    },
+    {
+      "i": 25,
+      "category": "prose",
+      "transcript": "Um but the extra step that we're asking you to do in this case is to look at the silent regions and listen to them for each channel and um make sure that no back channels were lost, for example.",
+      "target": "But the extra step that we're asking you to do in this case is to look at the silent regions and listen to them for each channel, and make sure that no back channels were lost, for example.",
+      "output": "But the extra step that we're asking you to do in this case is to look at the silent regions and listen to them for each channel, and make sure that no back channels were lost, for example.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 26,
+      "category": "meeting",
+      "transcript": "Are they doing um uh are are they uh as in the kickoff meeting just exchanging information to get started with the project or are they uh actually addressing some specific design um criterion at the moment?",
+      "target": "Are they doing, are they\u2014as in the kick-off meeting just exchanging information to get started with a project, or are they actually addressing some specific design criteria at the moment?",
+      "output": "Are they doing\u2014are they doing, as in the kickoff meeting, just exchanging information to get started with the project, or are they actually addressing some specific design criterion at the moment?",
+      "exact": false,
+      "similarity": 0.914,
+      "target_coverage": 0.742,
+      "forbidden": []
+    },
+    {
+      "i": 27,
+      "category": "prose",
+      "transcript": "Would we wanna ask them and put some kind of like proviso in to make it um specific to sort of preemptively",
+      "target": "Would we wanna ask them and put some kind of, like, provisor in to make it specific, to sort of preemptively...",
+      "output": "Would we wanna ask them and put some kind of, like, proviso, in, to make it specific to sort of preemptively?",
+      "exact": false,
+      "similarity": 0.964,
+      "target_coverage": 0.9,
+      "forbidden": []
+    },
+    {
+      "i": 28,
+      "category": "prose",
+      "transcript": "And there's especially one there that um I think he's Eastern European and I think that's the only reason I can understand what he's saying at all.",
+      "target": "And there's especially one there that, I think he's Eastern European, and I think that's the only reason I can understand what he's saying at all.",
+      "output": "And there's especially one there that\u2014I think he's Eastern European, and I think that's the only reason I can understand what he's saying at all.",
+      "exact": false,
+      "similarity": 0.99,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 29,
+      "category": "meeting",
+      "transcript": "Because um I I think it's actually very similar to what I did for my um speaker um uh extraction and I think I would perhaps have to change two lines of code to get you um for each meeting a file that says fr from um this millisecond to this millisecond there was this sequence of words.",
+      "target": "Because I think it's actually very similar to what I did for my speaker extraction, and I think I would perhaps have to change two lines of code to get you, for each meeting, a file that says from this millisecond to this millisecond there was this sequence of words.",
+      "output": "Because I think it's actually very similar to what I did for my speaker, extraction, and I would perhaps have to change two lines of code to get you for each meeting a file that says 'from this millisecond to this millisecond' there was this sequence of words.",
+      "exact": false,
+      "similarity": 0.421,
+      "target_coverage": 0.98,
+      "forbidden": []
+    },
+    {
+      "i": 30,
+      "category": "prose",
+      "transcript": "I would just try to keep segments very manageable. Um and but y you use these uh waveforms to help you.",
+      "target": "I would just try to keep segments very manageable, but, you know, use these waveforms to help you.",
+      "output": "I would just try to keep segments very manageable, and but you use these waveforms to help you.",
+      "exact": false,
+      "similarity": 0.943,
+      "target_coverage": 0.824,
+      "forbidden": []
+    },
+    {
+      "i": 31,
+      "category": "prose",
+      "transcript": "Um right into the call that we are considering the idea of having a poster session.",
+      "target": "Write into the call that we are considering the idea of having a poster session.",
+      "output": "Right into the call that we are considering the idea of having a poster session.",
+      "exact": false,
+      "similarity": 0.963,
+      "target_coverage": 0.929,
+      "forbidden": []
+    },
+    {
+      "i": 32,
+      "category": "prose",
+      "transcript": "Well I'm I'm I'm I'm trying tr I'm making sure that the GDA format we've got at the moment will will output the well output all the analysis that that Robin's looking for.",
+      "target": "Well, I'm just trying\u2014I'm making sure that the GDF format we've got at the moment will output\u2014say, well, I've output old analysis that Robin's looking for.",
+      "output": "Well, I'm\u2014I'm\u2014I'm\u2014I'm trying, making sure that the GDA format we've got at the moment will output the well output all the analysis that Robin's looking for.",
+      "exact": false,
+      "similarity": 0.849,
+      "target_coverage": 0.625,
+      "forbidden": []
+    },
+    {
+      "i": 33,
+      "category": "prose",
+      "transcript": "I can just add SSH stuff in to ev evoke the sound tools on the uh on dice and then copy them back.",
+      "target": "I can just add SSH stuff in to evoke the sound tools on Dice, and then copy 'em back.",
+      "output": "I can just add SSH stuff in to evoke the sound tools on the dice, then copy them back.",
+      "exact": false,
+      "similarity": 0.924,
+      "target_coverage": 0.722,
+      "forbidden": []
+    },
+    {
+      "i": 34,
+      "category": "prose",
+      "transcript": "have all the l lines of actually lots of different elements and you have to work out like all of the ones which are iron lines.",
+      "target": "You have all the lines of actually lots of different elements, and you have to work out at all the ones which are ion lines.",
+      "output": "Have all the L lines of actually lots of different elements, and you have to work out, like, all of the ones which are iron lines.",
+      "exact": false,
+      "similarity": 0.921,
+      "target_coverage": 0.792,
+      "forbidden": []
+    },
+    {
+      "i": 35,
+      "category": "prose",
+      "transcript": "Um suppose it doesn't really matter as long as they're unique and it can I can pass them back to you and you can use that to the display.",
+      "target": "I suppose it doesn't really matter as long as they're unique, and I can pass 'em back to you and you can use that to the display.",
+      "output": "Suppose it doesn't really matter as long as they're unique and it can\u2014I can pass them back to you, and you can use that to the display.",
+      "exact": false,
+      "similarity": 0.939,
+      "target_coverage": 0.857,
+      "forbidden": []
+    },
+    {
+      "i": 36,
+      "category": "prose",
+      "transcript": "So um I don't know, just just to to ask now if you've got anything else you've thought about while we've been talking.",
+      "target": "So, dunno, just to ask now if you've got anything else you've thought about while we've been talking.",
+      "output": "So I don't know, just to ask now if you've got anything else you've thought about while we've been talking.",
+      "exact": false,
+      "similarity": 0.952,
+      "target_coverage": 0.9,
+      "forbidden": []
+    },
+    {
+      "i": 37,
+      "category": "meeting",
+      "transcript": "I mean there probably are a bunch of issues like if you've got a hundred hours of meetings and they all talk about a lot of the same things.",
+      "target": "I mean, there probably are a bunch of issues, like, if you've got a hundred hours of meetings and they all talk about a lot of the same things.",
+      "output": "I mean, probably there are a bunch of issues like if you've got a hundred hours of meetings and they all talk about a lot of the same things.",
+      "exact": false,
+      "similarity": 0.951,
+      "target_coverage": 0.897,
+      "forbidden": []
+    },
+    {
+      "i": 38,
+      "category": "prose",
+      "transcript": "Um at least it works if there are several categories with each with one document each, but it um",
+      "target": "At least it works if there are several categories, with each, with one document each, but it\u2014yeah.",
+      "output": "At least it works if there are several categories with each with one document, but it...",
+      "exact": false,
+      "similarity": 0.925,
+      "target_coverage": 0.824,
+      "forbidden": []
+    },
+    {
+      "i": 39,
+      "category": "prose",
+      "transcript": "Suppose we pick up an we we we screwed up the first time we put the triangle with the square and we threw it away. We decided we didn't like it.",
+      "target": "Suppose we screwed up the first time we put the triangle with the square, and we threw it away and decided we didn't like it.",
+      "output": "Suppose we pick up\u2014we screwed up the first time we put the triangle with the square, and we threw it away. We decided we didn't like it.",
+      "exact": false,
+      "similarity": 0.935,
+      "target_coverage": 0.92,
+      "forbidden": []
+    },
+    {
+      "i": 40,
+      "category": "prose",
+      "transcript": "I think proportional like I even though I'm I'm always getting myself confused about this, but I think sort of making the summary length thing generally proportional",
+      "target": "I think proportion, like, even though I'm always getting myself confused about this, but I think sort of making the summary length thing generally proportion.",
+      "output": "I think proportional, like, I even though I'm always getting myself confused about this, but I think sort of making the summary length thing generally proportional.",
+      "exact": false,
+      "similarity": 0.981,
+      "target_coverage": 0.84,
+      "forbidden": []
+    },
+    {
+      "i": 41,
+      "category": "prose",
+      "transcript": "Uh what el uh well I'll I'll have to get to grips with Cra Camtasia uh properly. Uh to generate a little sort of introductory video, I suppose, uh for instructional purposes at the beginning of the experiment.",
+      "target": "Well, I'll have to get to grips with Camtasia properly to generate a little sort of introductory video, I suppose, for instructional purposes at the beginning of the experiment.",
+      "output": "Well, I'll have to get to grips with CraCamTasia properly. To generate a little sort of introductory video, I suppose, for instructional purposes at the beginning of the experiment.",
+      "exact": false,
+      "similarity": 0.978,
+      "target_coverage": 0.931,
+      "forbidden": []
+    },
+    {
+      "i": 42,
+      "category": "prose",
+      "transcript": "Um what I because nine megabyte is what I got for when I said for every um utterance. This goes from there to there and",
+      "target": "What I\u2014because nine megabytes is what I got for when I said for every utterance, this goes from there to there, and...",
+      "output": "What I\u2014because nine megabytes is what I got for when I said for every utterance, this goes from there to there.",
+      "exact": false,
+      "similarity": 0.969,
+      "target_coverage": 0.955,
+      "forbidden": []
+    },
+    {
+      "i": 43,
+      "category": "prose",
+      "transcript": "If via the wasn't in there then you'd have it silicon two lines at but without with via the I think that means silicon two derived from these two features.",
+      "target": "If it wasn't in there, then you'd have it 'silicone two lines at' but without\u2014with the\u2014I think that means silicone two derived from these two features.",
+      "output": "If via the wasn't in there, then you'd have it silicon two lines at, but without via the\u2014I think that means silicon two derived from these two features.",
+      "exact": false,
+      "similarity": 0.944,
+      "target_coverage": 0.724,
+      "forbidden": []
+    },
+    {
+      "i": 44,
+      "category": "prose",
+      "transcript": "So what I've been thinking is maybe if we try to really make sense together of the XML format to be sure that we can all produce the data that we are producing in a way or at least I mean, I guess my data I'll probably load in from completely different way anyway because it's a matrix.",
+      "target": "So what I've been thinking is, maybe if we try to really make sense together of the XML format to be sure that we can all produce the data that we are producing in a way\u2014or at least, I mean, I guess my data I'll probably load in from completely different way anyway because it's matrix.",
+      "output": "So what I've been thinking is, maybe if we try to really make sense together of the XML format, to be sure that we can all produce the data that we are producing in a way, or at least\u2014I mean, my data, I'll probably load in from completely different way anyway, because it's a matrix.",
+      "exact": false,
+      "similarity": 0.626,
+      "target_coverage": 0.932,
+      "forbidden": []
+    },
+    {
+      "i": 45,
+      "category": "prose",
+      "transcript": "Mm. Cause uh so for the information and action the uh the action here is not direct but uh some kind of uh uh physical action uh that uh the uh participants uh i is going to uh or has done in the past.",
+      "target": "'Cause, so, for the information and action, the action here is not dialogue act but some kind of physical action that the participant is going to or has done in the past.",
+      "output": "Cause, so for the information and action, the action here is not direct but some kind of physical action, that the participants will have, or has done in the past.",
+      "exact": false,
+      "similarity": 0.919,
+      "target_coverage": 0.742,
+      "forbidden": [
+        "here is"
+      ]
+    },
+    {
+      "i": 46,
+      "category": "meeting",
+      "transcript": "J because uh after talking to Jim, um just uh we have different ideas whether um meeting action will be about the same size as topic segments.",
+      "target": "Because after talking to Jane, just\u2014we have different ideas whether meeting action will be about the same size as topic segments.",
+      "output": "J because after talking to Jim, just we have different ideas whether meeting action will be about the same size as topic segments.",
+      "exact": false,
+      "similarity": 0.958,
+      "target_coverage": 0.905,
+      "forbidden": []
+    },
+    {
+      "i": 47,
+      "category": "prose",
+      "transcript": "uh i it's also going to be a problem, uh, because you're going to have one part or construct forming in the middle of a fixation.",
+      "target": "It's also going to be a problem because you're going to have one part or a construct forming in the middle of a fixation.",
+      "output": "It's also going to be a problem, because you're going to have one part or construct forming in the middle of a fixation.",
+      "exact": false,
+      "similarity": 0.988,
+      "target_coverage": 0.92,
+      "forbidden": []
+    },
+    {
+      "i": 48,
+      "category": "prose",
+      "transcript": "Um it ties you down perhaps a bit later on in terms of the technologies, um, how far you can transmit a signal or the complexity of the functions that you want.",
+      "target": "It ties you down perhaps a bit later on in terms of the technologies, how far you can transmit the signal, or the complexity of the functions that you want.",
+      "output": "It ties you down perhaps a bit later on in terms of the technologies, how far you can transmit a signal, or the complexity of the functions that you want.",
+      "exact": false,
+      "similarity": 0.987,
+      "target_coverage": 0.931,
+      "forbidden": []
+    },
+    {
+      "i": 49,
+      "category": "prose",
+      "transcript": "Like that initially had quite different ideas about the phrase types and and markations and then eventually you came up with some kind of rules that you set up between uh the two of you.",
+      "target": "Like, that initially had quite different ideas about the phrase types and markations, and then eventually you came up with some kind of rules that you set up between the two of you.",
+      "output": "Like, that initially had quite different ideas about the phrase types and markings, and then eventually you came up with some kind of rules that you set up between the two of you.",
+      "exact": false,
+      "similarity": 0.989,
+      "target_coverage": 0.938,
+      "forbidden": []
+    },
+    {
+      "i": 50,
+      "category": "meeting",
+      "transcript": "And so I think going through the presentations that um were given at the work package six meeting where um I think it was um it wasn't w was it Anita or was it oh I've forgotten her name.",
+      "target": "And so I think going through the presentations that were given at the work package six meeting, where, and I think it was\u2014it wasn't\u2014was it Anita, or was it\u2014oh, I've forgotten her name.",
+      "output": "And so I think going through the presentations that were given at the work package 6 meeting where I think it was\u2014was it\u2014anita or was it, oh I've forgotten her name.",
+      "exact": false,
+      "similarity": 0.917,
+      "target_coverage": 0.811,
+      "forbidden": []
+    },
+    {
+      "i": 51,
+      "category": "prose",
+      "transcript": "Um frequencies or something. Um as well as um characters, um different uh keypad styles and symbols.",
+      "target": "Frequencies or something, as well as characters, different keypad styles and symbols.",
+      "output": "Frequencies or something, as well as characters, different keypad styles and symbols.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 52,
+      "category": "prose",
+      "transcript": "a couple of hours getting photocopies of degree certificates to give you for your files if I'm just gonna be here'cause they want I think it that one place it said give us copies of degree",
+      "target": "A couple of hours getting photocopies of degree certificates to give you for your files, if I'm just gonna be here a good\u2014they won't, I think at one place it said give us copies of degrees.",
+      "output": "A couple of hours getting photocopies of degree certificates to give you for your files, if I'm just gonna be here, 'cause they want\u2014I think that one place said give us copies of degree.",
+      "exact": false,
+      "similarity": 0.939,
+      "target_coverage": 0.711,
+      "forbidden": []
+    },
+    {
+      "i": 53,
+      "category": "prose",
+      "transcript": "What um the kinds of events that we had uh before talked about putting into the record for then use in Alan and NXT isn't based on saying every X second something",
+      "target": "What\u2014the kinds of events that we had before talked about putting into the record for then use in ELAN and NXT isn't based on saying every X seconds something.",
+      "output": "What the kinds of events that we had before, talked about, putting into the record for then, use in Alan and NXT, isn't based on saying every X second something.",
+      "exact": false,
+      "similarity": 0.959,
+      "target_coverage": 0.867,
+      "forbidden": []
+    },
+    {
+      "i": 54,
+      "category": "prose",
+      "transcript": "Mm we were doing descriptions that made things separate classes, in which case I'd argue for elliptical, I think.",
+      "target": "We were doing descriptions that made things separate classes, in which case I'd argue for elliptical, I think.",
+      "output": "We were doing descriptions that made things separate classes, in which case I'd argue for elliptical, I think.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 55,
+      "category": "meeting",
+      "transcript": "So uh in meetings I uh from my uh reading of the one of the M meetings IS one oh eight uh the uh the four meetings",
+      "target": "So in meetings, from my reading of one of the AMI meetings, IS one O eight, the four meetings.",
+      "output": "So in meetings I, from my reading of the one of the meetings, is one eight the four meetings.",
+      "exact": false,
+      "similarity": 0.909,
+      "target_coverage": 0.667,
+      "forbidden": []
+    },
+    {
+      "i": 56,
+      "category": "meeting",
+      "transcript": "And then between us we'll have to have another meeting and and I mean, at the same time as we're sorting through all the other titles and and deciding um how to divide up the sessions.",
+      "target": "And then between us we'll have to have another meeting, and I mean, at the same time as we're sorting through all the other titles and deciding how to divide up the sessions.",
+      "output": "And then between us we'll have to have another meeting, and I mean, at the same time as we're sorting through all the other titles and deciding how to divide up the sessions.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 57,
+      "category": "prose",
+      "transcript": "So this person left out a lot of N punctuation and um here's a case where I think she was transcribing an outbreath, um, or possibly an in breath, I uh in and used the hash mark.",
+      "target": "So this person left out a lot of end punctuation, and here's a case where I think she was transcribing an out-breath, or possibly an in-breath, in, and used the hash mark.",
+      "output": "So this person left out a lot of punctuation, and here's a case where I think she was transcribing an utterance, or possibly an in-brattle, I in, and used the hash mark.",
+      "exact": false,
+      "similarity": 0.929,
+      "target_coverage": 0.794,
+      "forbidden": []
+    },
+    {
+      "i": 58,
+      "category": "prose",
+      "transcript": "And I actually uh think that that's a slightly extreme um position um with regard to the literature that a lot of people might say it doesn't really matter who the different speakers are in an interactive section.",
+      "target": "And I actually think that that's a slightly extreme position with regard to the literature, that a lot of people might say it doesn't really matter who the different speakers are in an interactive section.",
+      "output": "And I actually think that's a slightly extreme position with regard to the literature that a lot of people might say it doesn't really matter who the different speakers are in an interactive section.",
+      "exact": false,
+      "similarity": 0.985,
+      "target_coverage": 0.972,
+      "forbidden": []
+    },
+    {
+      "i": 59,
+      "category": "prose",
+      "transcript": "Oh, I had one guy, one of the project managers was um he had a really weird accent and he was calling um one guy Mr. Red.",
+      "target": "Oh, I had one guy\u2014one of the project managers was, he had a really weird accent and he was calling one guy 'Mister Ed'.",
+      "output": "Oh, I had one guy, one of the project managers, he had a really weird accent, and he was calling one guy Mr. Red.",
+      "exact": false,
+      "similarity": 0.905,
+      "target_coverage": 0.833,
+      "forbidden": []
+    },
+    {
+      "i": 60,
+      "category": "prose",
+      "transcript": "By now I just had a look at the year data. Um and actually the average F zero deviation is only slightly higher for the positive ones than for the negative ones.",
+      "target": "By now, I just had a look at the data, and actually the average F-zero deviation is only slightly higher for the positive ones than for the negative ones.",
+      "output": "By now I just had a look at the year data, and actually the average F-zero deviation is only slightly higher for the positive ones than for the negative ones.",
+      "exact": false,
+      "similarity": 0.981,
+      "target_coverage": 0.966,
+      "forbidden": []
+    },
+    {
+      "i": 61,
+      "category": "prose",
+      "transcript": "Um I would recommend I've found that it's easier when you're checking and you're trying to make sure the boundaries are right for a speech segment.",
+      "target": "I would recommend\u2014I've found that it's easier when you're checking and you're trying to make sure the boundaries are right for a speech segment.",
+      "output": "I would recommend I've found that it's easier when you're checking, and you're trying to make sure the boundaries are right for a speech segment.",
+      "exact": false,
+      "similarity": 0.99,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 62,
+      "category": "prose",
+      "transcript": "But the the text for the actual call, is she going to rework something or uh I think she should",
+      "target": "But the text for the actual call, is she going to rework something, or she should?",
+      "output": "But the text for the actual call is she going to rework something, or\u2014I think she should.",
+      "exact": false,
+      "similarity": 0.936,
+      "target_coverage": 0.933,
+      "forbidden": []
+    },
+    {
+      "i": 63,
+      "category": "prose",
+      "transcript": "We've got we've got all the data in the ASCII format at the moment, and it's just a matter of what data do we need to pull out of that into the GDF so I can actually do the prop analysis.",
+      "target": "'Cause we got all this in the ASCII format at the moment, and it's just a matter of what data do we need to pull out of that into the GDF, so I can actually do the proper analysis.",
+      "output": "We've got all the data in the ASCII format at the moment, and it's just a matter of what data we need to pull out of that into the GDF so I can actually do the prop analysis.",
+      "exact": false,
+      "similarity": 0.932,
+      "target_coverage": 0.795,
+      "forbidden": []
+    },
+    {
+      "i": 64,
+      "category": "prose",
+      "transcript": "I think it's'cause we had to specify it ourselves that it's not as um like focus a specification of most um work we have to do.",
+      "target": "I think it's 'cause we had to specify it ourselves, that it's not as, like, focus the specification of most work we have to do.",
+      "output": "I think it's 'cause we had to specify it ourselves that it's not as, like, focus specification of most work we have to do.",
+      "exact": false,
+      "similarity": 0.98,
+      "target_coverage": 0.923,
+      "forbidden": []
+    },
+    {
+      "i": 65,
+      "category": "prose",
+      "transcript": "Yeah, Dave C in the program manager, do you have a a method to like set the current ID so it'll update the display?",
+      "target": "Yeah, Dave, see, in the programme manager, do you have a method to, like, set the current ID so it'll update the display?",
+      "output": "Yeah, Dave in the programme manager, do you have a method to, like, set the current ID so it'll update the display?",
+      "exact": false,
+      "similarity": 0.975,
+      "target_coverage": 0.913,
+      "forbidden": []
+    },
+    {
+      "i": 66,
+      "category": "prose",
+      "transcript": "um which is the the first one is that um uh the companies decided that teletext is outdated uh because of how popular the internet is.",
+      "target": "Which is the first one, is that the company's decided that teletext is outdated because of how popular the internet is.",
+      "output": "Which is the first one, that the companies decided that teletext is outdated because of how popular the internet is?",
+      "exact": false,
+      "similarity": 0.962,
+      "target_coverage": 0.762,
+      "forbidden": []
+    },
+    {
+      "i": 67,
+      "category": "meeting",
+      "transcript": "So should we expect that all the slides from like the work package six meeting or the joint work package five work pack will be up on one of the work package websites?",
+      "target": "So should we expect that all the slides from, like, the work package six meeting or the joint work package five work package will be up on one of the work package websites?",
+      "output": "So should we expect that all the slides from, like the work package six meeting, or the joint work package five work pack, will be up on one of the work package websites?",
+      "exact": false,
+      "similarity": 0.982,
+      "target_coverage": 0.969,
+      "forbidden": []
+    },
+    {
+      "i": 68,
+      "category": "prose",
+      "transcript": "If we need just frequencies, maybe we should just calculate them by using Pearl or something.",
+      "target": "If we need just frequencies, maybe we should just calculate them by using Perl or something.",
+      "output": "If we need just frequencies, maybe we should just calculate them by using Pearl or something.",
+      "exact": false,
+      "similarity": 0.995,
+      "target_coverage": 0.867,
+      "forbidden": []
+    },
+    {
+      "i": 69,
+      "category": "prose",
+      "transcript": "So so essentially there's this um A E is divided into T R one and C and other stuff. Okay, where C is the construct, the existing thing.",
+      "target": "So essentially there's a AE, is divided into TR-one and C, and other stuff, okay, where C is the construct, the existing thing.",
+      "output": "So essentially there's this A-E is divided into T-R-one and C, and other stuff, okay, where C is the construct, the existing thing.",
+      "exact": false,
+      "similarity": 0.969,
+      "target_coverage": 0.792,
+      "forbidden": []
+    },
+    {
+      "i": 70,
+      "category": "prose",
+      "transcript": "If I have to do it f like flexibly then I probably then it'll not be significantly more difficult.",
+      "target": "If I have to do it flexibly, then I probably\u2014then it will not be significantly more difficult.",
+      "output": "If I have to do it, like, flexibly, then I probably\u2014then it'll not be significantly more difficult.",
+      "exact": false,
+      "similarity": 0.943,
+      "target_coverage": 0.824,
+      "forbidden": []
+    },
+    {
+      "i": 71,
+      "category": "prose",
+      "transcript": "And I've uh sent a thing off to Marlo's talk trying to encourage her again to register for the C S V and uh and what to C V S and uh wh what to do and uh which bit of software I use that seems to be quite easy. I was just imagining that okay, and you never",
+      "target": "And I've sent a thing off to Marloes Talk trying to encourage her again to register for the CVS, and what to\u2014the CVS, and what to do, and which bit of software I use, that seems to be quite easy and simple.",
+      "output": "And I've sent a thing off to Marlo's talk trying to encourage her again to register for the CSV and what to do and which bit of software I use that seems to be quite easy. I was just imagining that, okay, and you never.",
+      "exact": false,
+      "similarity": 0.715,
+      "target_coverage": 0.791,
+      "forbidden": []
+    },
+    {
+      "i": 72,
+      "category": "prose",
+      "transcript": "Yeah, what I did at the moment is um I um yeah, I got the F zero values from for each speaker for his headphone.",
+      "target": "Yeah, what I did at the moment is I got the F-zero values from\u2014for each speaker, for his headphone.",
+      "output": "Yeah, what I did at the moment is\u2014I, yeah, I got the F-zero values from for each speaker for his headphone.",
+      "exact": false,
+      "similarity": 0.942,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 73,
+      "category": "prose",
+      "transcript": "Oh gosh, narrow emission Okay, so I've highlighted this in red and then the whole thing in green because type two C foot galaxies are one.",
+      "target": "Oh gosh, no emission line, okay, so I've highlighted this in red, and then the whole thing in green because type two Seyfart galaxies are one.",
+      "output": "Oh gosh, narrow emission, okay, so I've highlighted this in red, and then the whole thing in green, because type 2C galaxies are one.",
+      "exact": false,
+      "similarity": 0.916,
+      "target_coverage": 0.731,
+      "forbidden": []
+    },
+    {
+      "i": 74,
+      "category": "meeting",
+      "transcript": "It more and more appears to me that if we if we scrap the notion of the meeting as an individual thing and sort of ch see meetings as as topic segments and have sort of like hierarchical topic segmentation instead.",
+      "target": "It more and more appears to me that if we scrap the notion of the meeting as an individual thing and sort of just see meetings as topic segments and have sort of, like, hierarchical topic segmentation instead.",
+      "output": "It more and more appears to me that if we scrap the notion of the meeting as an individual thing and sort of\u2014ch-see meetings as topic segments and have sort of like hierarchical topic segmentation instead.",
+      "exact": false,
+      "similarity": 0.971,
+      "target_coverage": 0.946,
+      "forbidden": []
+    },
+    {
+      "i": 75,
+      "category": "prose",
+      "transcript": "So from uh um my analysis of that kind of dialogues, I could say that there's only there are only four basic uh exchanges depend on the information uh direction of information flow.",
+      "target": "So from my analysis of that kind of dialogues, I could say that there are only four basic exchanges, depending on the direction of the information flow.",
+      "output": "So from my analysis of that kind of dialogues, I could say that there's only\u2014there are only four basic exchanges, depend on the information direction of information flow.",
+      "exact": false,
+      "similarity": 0.901,
+      "target_coverage": 0.846,
+      "forbidden": []
+    },
+    {
+      "i": 76,
+      "category": "meeting",
+      "transcript": "I think they are more talking about um user requirement for the meeting than subtext they want to know.",
+      "target": "I think they are more talking about user requirement for the meeting than soft text they want to know.",
+      "output": "I think they are more talking about user requirement for the meeting than subtext they want to know.",
+      "exact": false,
+      "similarity": 0.97,
+      "target_coverage": 0.833,
+      "forbidden": []
+    },
+    {
+      "i": 77,
+      "category": "prose",
+      "transcript": "Um I've sorted out uh travel, etcetera, to Munich for uh for Ellen, John and myself.",
+      "target": "I've sorted out travel et cetera to Munich for Ellen, John, and myself.",
+      "output": "I've sorted out travel, etcetera, to Munich for Ellen, John, and myself.",
+      "exact": false,
+      "similarity": 0.979,
+      "target_coverage": 0.769,
+      "forbidden": []
+    },
+    {
+      "i": 78,
+      "category": "prose",
+      "transcript": "It does make sense from maybe the design point of view'cause if you have more complicated characters like European languages then you need more buttons.",
+      "target": "It does make sense from maybe the design point of view, 'cause you have more complicated characters, like European languages, then you need more buttons.",
+      "output": "It does make sense from maybe the design point of view, 'cause if you have more complicated characters, like European languages, then you need more buttons.",
+      "exact": false,
+      "similarity": 0.99,
+      "target_coverage": 0.958,
+      "forbidden": []
+    },
+    {
+      "i": 79,
+      "category": "prose",
+      "transcript": "So it's more like a it's uh it's sort of like a quite literal task. So it's really like finding these kind of uh these types of strings uh rather than really referring to the underlying physical reality.",
+      "target": "So it's more like\u2014it's sort of like a quite literal task that's really like finding these kind of, these types of strings, rather than really referring to the underlying physical reality.",
+      "output": "So it's more like, sort of, like, a quite literal task, so it's really, like, finding these kind of\u2014these types of strings rather than really referring to the underlying physical reality.",
+      "exact": false,
+      "similarity": 0.941,
+      "target_coverage": 0.882,
+      "forbidden": []
+    },
+    {
+      "i": 80,
+      "category": "prose",
+      "transcript": "And yeah, that they have lots of personality and uh be fit and in robust good health.",
+      "target": "And yeah, that they have lots of personality, and be fit, and in robust good health.",
+      "output": "And yeah, that they have lots of personality, and be fit and in robust good health.",
+      "exact": false,
+      "similarity": 0.994,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 81,
+      "category": "prose",
+      "transcript": "And sometimes with the ds uh if you're thinking, are you anticipating a d or you're you trying to say th and you're just saying the the the the the",
+      "target": "And sometimes with the\u2014I would be thinking, are you anticipating, or are you trying to say 'the' and you're just saying 'the the the the the'?",
+      "output": "And sometimes with the\u2014if you're thinking, are you anticipating a\u2014you're trying to say, you're just saying the\u2014the, the, the, the...",
+      "exact": false,
+      "similarity": 0.818,
+      "target_coverage": 0.63,
+      "forbidden": []
+    },
+    {
+      "i": 82,
+      "category": "prose",
+      "transcript": "So I mean I I do think you should keep in mind what the final goal is so that you can plug this into something at the end and and um show some benefit of having the markup.",
+      "target": "So I mean, I do think you should keep in mind what the final goal is, so that you can plug this into something at the end and show some benefit of having the markup.",
+      "output": "So I mean, I do think you should keep in mind what the final goal is, so that you can plug this into something at the end and show some benefit of having the markup.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 83,
+      "category": "prose",
+      "transcript": "But you wouldn't be able to point to something in a spectrum and say that's a damped Lyman Alpha Absorber unless you'd",
+      "target": "But you wouldn't be able to point something in a spectrum and say that's a damped Lyman alpha absorber unless you'd...",
+      "output": "But you wouldn't be able to point to something in a spectrum and say that's a damped Lyman alpha absorber unless you'd.",
+      "exact": false,
+      "similarity": 0.979,
+      "target_coverage": 0.957,
+      "forbidden": []
+    },
+    {
+      "i": 84,
+      "category": "meeting",
+      "transcript": "So uh and currently I'm more interested in the uh uh in analyzing the disconstruction multi party meetings.",
+      "target": "So, currently I'm more interested in, yeah, analysing the discourse structure in multi-party meetings.",
+      "output": "So currently I'm more interested in analysing the disconstruction multi-party meetings.",
+      "exact": false,
+      "similarity": 0.899,
+      "target_coverage": 0.6,
+      "forbidden": []
+    },
+    {
+      "i": 85,
+      "category": "prose",
+      "transcript": "In fact, even by inviting um people from um informatics and from well maybe from philosophy and things, it might bias it in a particular direction.",
+      "target": "In fact, even by inviting people from Informatics and from\u2014or maybe from Philosophy and things\u2014it might bias it in a particular direction.",
+      "output": "In fact, even by inviting people from informatics and from philosophy, it might bias it in a particular direction.",
+      "exact": false,
+      "similarity": 0.881,
+      "target_coverage": 0.739,
+      "forbidden": []
+    },
+    {
+      "i": 86,
+      "category": "prose",
+      "transcript": "But they're they're creating a there. Uh guy, Jonathan, is creating a web form for you to do this all very simply and like just um upload everything that you've uh uh upload the file that you've transcribed and and you'll be noting whether it's the first pass or the second pass and your name um as the person who did that particular pass.",
+      "target": "But they're creating\u2014their guy Jonathan is creating a web form for you to do this all very simply, and like just upload everything that you've\u2014upload the file that you've transcribed, and you'll be noting whether it's the first pass or the second pass, and your name as the person who did that particular pass.",
+      "output": "But they're creating a\u2014there, guy, Jonathan, is creating a web form for you to do this all very simple, and, like, just upload everything that you've transcribed, and you'll be noting whether it's the first pass or the second pass, and your name as the person who did that particular pass.",
+      "exact": false,
+      "similarity": 0.2,
+      "target_coverage": 0.867,
+      "forbidden": []
+    },
+    {
+      "i": 87,
+      "category": "prose",
+      "transcript": "So one speaker will have the floored channel um throughout that. And so I guess what you're saying is that the interactive sections would have shorter speaker turns when you're not considering the back channels as interruptions.",
+      "target": "So one speaker will have the floor channel throughout that, and so I guess what you're saying is that the interactive sections would have shorter speaker turns when you're not considering the back channels as interruptions.",
+      "output": "So one speaker will have the flopped channel throughout that, and so I guess what you're saying is that the interactive sections would have shorter speaker turns when you're not considering the back channels as interruptions.",
+      "exact": false,
+      "similarity": 0.179,
+      "target_coverage": 0.946,
+      "forbidden": []
+    },
+    {
+      "i": 88,
+      "category": "prose",
+      "transcript": "It's usually a context problem because I think sometimes if it's not intelligible, you either tend to because you really want it to fit the context",
+      "target": "It's usually a context problem, because I think sometimes if it's not intelligible, you either tend to\u2014because you really want it to fit the context.",
+      "output": "It's usually a context problem because I think sometimes if it's not intelligible, you either tend to\u2014because you really want it to fit the context.",
+      "exact": false,
+      "similarity": 0.997,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 89,
+      "category": "meeting",
+      "transcript": "Couldn't we couldn't we write one program that calculates all the um f for all our different contributions to the importance measure the um absolute um dialogue act numbers for the meeting?",
+      "target": "Couldn't we write one programme that calculates all the\u2014for all our different contributions to the importance measure, the absolute dialogue act numbers for the meeting?",
+      "output": "Couldn't we write one programme that calculates all the f for all our different contributions to the importance measure, the absolute dialogue act numbers for the meeting?",
+      "exact": false,
+      "similarity": 0.988,
+      "target_coverage": 0.962,
+      "forbidden": [
+        "sure,"
+      ]
+    },
+    {
+      "i": 90,
+      "category": "prose",
+      "transcript": "So no um n no transcriber will be checking no checker will be checking their own transcription.",
+      "target": "So no transcriber will be checking\u2014no checker will be checking their own transcription.",
+      "output": "Transcriber will be checking their own transcription.",
+      "exact": false,
+      "similarity": 0.743,
+      "target_coverage": 0.615,
+      "forbidden": []
+    },
+    {
+      "i": 91,
+      "category": "prose",
+      "transcript": "In some cases we might if it's really tough, we might be in need of a of some sound system to play recordings.",
+      "target": "In some cases we might\u2014if it's really tough we might be in need of some sound system to play recordings.",
+      "output": "In some cases we might\u2014if it's really tough\u2014we might be in need of some sound system to play recordings.",
+      "exact": false,
+      "similarity": 0.99,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 92,
+      "category": "prose",
+      "transcript": "So at the moment I'm looking at act actually the GDF format itself. Um and then to generate the GDA format I'm gonna be taking stuff out of the Jest analyzer um and then using the DOM parser to generate uh XML output from the Gest analyzer.",
+      "target": "So at the moment I'm looking at actually the GDF format itself, and then to generate the GDF format I'm gonna be taking stuff out of the gest analyser, and then using a DOM parser to generate XML output from the gest analyser.",
+      "output": "So at the moment I'm looking at\u2014actually, the GDF format itself, and then to generate the GDA format I'm gonna be taking stuff out of the Jest analyzer, and then using the DOM parser to generate XML output from the Gest analyzer.",
+      "exact": false,
+      "similarity": 0.963,
+      "target_coverage": 0.886,
+      "forbidden": []
+    },
+    {
+      "i": 93,
+      "category": "prose",
+      "transcript": "Well what I think I might try and build is basically a class that you just feed it a linked list of um different waveforms.",
+      "target": "Well, what I think I might try and build is basically a class that you just feed it a linked list of different waveforms.",
+      "output": "Well, what I think I might try and build is basically a class that you just feed it a linked list of different waveforms.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 94,
+      "category": "intent",
+      "transcript": "So I guess um if I'm gonna be segmenting it with a L C Seg then that's the same format I'd want to um put it back out in, so it'd be equivalent.",
+      "target": "So I guess if I'm gonna be segmenting it with an LCSeg, then that's, like, same format I'd want to put it back out in, so it'd be equivalent.",
+      "output": "So I guess, if I'm gonna be segmenting it with a LCS, then that's the same format\u2014I'd want to put it back out in, so it'd be equivalent.",
+      "exact": false,
+      "similarity": 0.953,
+      "target_coverage": 0.844,
+      "forbidden": []
+    },
+    {
+      "i": 95,
+      "category": "prose",
+      "transcript": "Well well a worry that was was expressed in the new requirements was that if we made this too complex it would it it would affect um how long how long it took us to get this to market.",
+      "target": "Well, a worry that was expressed in the new requirements was that, if we made this too complex, it would affect how long it took us to get this to market.",
+      "output": "Well, a worry that was expressed in the new requirements, was that if we made this too complex it would affect how long it took us to get this to market.",
+      "exact": false,
+      "similarity": 0.99,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 96,
+      "category": "prose",
+      "transcript": "Um and so um it is important that these relations have a semantics that's tied to the semantics of the utterances, right?",
+      "target": "And so it is important that these relations have a semantics that's tied to the semantics of the utterances, right?",
+      "output": "And so it is important that these relations have a semantics that's tied to the semantics of the utterances, right?",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 97,
+      "category": "prose",
+      "transcript": "Yeah, that's what I thought as well. That you that probably the the topic segment level is the most um informative for the words.",
+      "target": "Yeah, that's what I thought as well, that probably the topic segment level is the most informative for the words.",
+      "output": "Yeah, that's what I thought, you that probably the topic segment level is the most informative for the words.",
+      "exact": false,
+      "similarity": 0.946,
+      "target_coverage": 0.85,
+      "forbidden": []
+    },
+    {
+      "i": 98,
+      "category": "prose",
+      "transcript": "If people are busy interacting all the time and all of our colleagues don't even bother to control for whether people are talking to one another when they're doing these joint tasks because it seems to them to be irrelevant because language is irrelevant.",
+      "target": "If people are busy interacting all the time and all of our colleagues don't even bother to control for whether people are talking to one another when they're doing these joint tasks, because it seems to them to be irrelevant, 'cause language is irrelevant.",
+      "output": "If people are busy interacting all the time and all of our colleagues don't even bother to control for whether people are talking to one another when they're doing these joint tasks, because it seems to them that it's irrelevant because language is irrelevant.",
+      "exact": false,
+      "similarity": 0.864,
+      "target_coverage": 0.889,
+      "forbidden": []
+    },
+    {
+      "i": 99,
+      "category": "prose",
+      "transcript": "It's sort of it's overrunning the f like it. So okay, so at the moment we are that far that if I now click onto a segment here it's",
+      "target": "It's sort of\u2014it's overrunning the F, like, it\u2014so okay, so at the moment we're that far that if I now click onto a segment, it hits.",
+      "output": "It's sort of\u2014overrunning the f, like, it, so okay, so at the moment we are that far, if I now click onto a segment here, it's...",
+      "exact": false,
+      "similarity": 0.88,
+      "target_coverage": 0.733,
+      "forbidden": []
+    }
+  ]
+}

eval/eval_ami_v2_1_mlx_4w_g32.json

@@ -0,0 +1,1145 @@
+{
+  "overall": {
+    "n": 100,
+    "forbidden_rate": 0.01,
+    "exact_match_rate": 0.03,
+    "mean_similarity": 0.6910109849919605,
+    "target_coverage": 0.5906611108170149,
+    "gates": {
+      "forbidden": true,
+      "coverage": false
+    },
+    "mean_tokens_per_sec": 476.51,
+    "verdict": "FAIL"
+  },
+  "by_category": {
+    "prose": {
+      "n": 83,
+      "forbidden_rate": 0.012,
+      "exact_match_rate": 0.0361,
+      "mean_similarity": 0.7022111087107978,
+      "target_coverage": 0.6126178160830319
+    },
+    "meeting": {
+      "n": 14,
+      "forbidden_rate": 0.0,
+      "exact_match_rate": 0.0,
+      "mean_similarity": 0.6459500827682932,
+      "target_coverage": 0.4985009530203537
+    },
+    "intent": {
+      "n": 3,
+      "forbidden_rate": 0.0,
+      "exact_match_rate": 0.0,
+      "mean_similarity": 0.5914251058145765,
+      "target_coverage": 0.41327300150829566
+    }
+  },
+  "forbidden_hits": {
+    "here's the": 1
+  },
+  "rows": [
+    {
+      "i": 0,
+      "category": "prose",
+      "transcript": "Oh, I think it is uh um s starting with um y European, I think.",
+      "target": "Oh, I think it's starting with European, I think.",
+      "output": "Starting with Y European, I think.",
+      "exact": false,
+      "similarity": 0.747,
+      "target_coverage": 0.444,
+      "forbidden": []
+    },
+    {
+      "i": 1,
+      "category": "meeting",
+      "transcript": "You could just like sort of zoom back out till you've got the full text for that segment or that meeting rather than a summary.",
+      "target": "You could just, like, sort of zoom back out so you got the full text for that segment or that meeting rather than a summary.",
+      "output": "You could just like sort of zoom back out till you've got the full text for that segment, or that meeting.",
+      "exact": false,
+      "similarity": 0.852,
+      "target_coverage": 0.708,
+      "forbidden": []
+    },
+    {
+      "i": 2,
+      "category": "prose",
+      "transcript": "Um so uh I guess just to confirm that we know what we're doing in the next well, in the thirty minutes after lunch anyway.",
+      "target": "So just to confirm that we know what we're doing in the\u2014well, in the thirty minutes after lunch anyway.",
+      "output": "Just to confirm that we know what we're doing in the next 30 minutes after lunch, anyway.",
+      "exact": false,
+      "similarity": 0.865,
+      "target_coverage": 0.75,
+      "forbidden": []
+    },
+    {
+      "i": 3,
+      "category": "intent",
+      "transcript": "And you might want to talk to Alex to find out if somebody's already done this turn redwood stuff into tree bank stuff so that they can compare",
+      "target": "And you might wanna talk to Alex to find out if somebody's already done this, turn Redwoods stuff into Treebank stuff, so that they can compare.",
+      "output": "And you might want to talk to Alex to find out if somebody's already done this turn redwood stuff into tree bank stuff, so that they can compare.",
+      "exact": false,
+      "similarity": 0.955,
+      "target_coverage": 0.769,
+      "forbidden": []
+    },
+    {
+      "i": 4,
+      "category": "prose",
+      "transcript": "Because it's quite easy in Pearl as well. It's just a line of code for counting all the words and Yeah, it's it's by hashes.",
+      "target": "Because it's quite easy in Perl as well, it's just a line of code for counting all the words, and yeah, it's by hashes.",
+      "output": "Because it's quite easy in Pearl, as well. It's just a line of code for counting all the words.",
+      "exact": false,
+      "similarity": 0.85,
+      "target_coverage": 0.731,
+      "forbidden": []
+    },
+    {
+      "i": 5,
+      "category": "prose",
+      "transcript": "If if your intention is to move the red triangle to sit on top of the thing already constructed, you would tend to look back and forth.",
+      "target": "If your intention is to move the red triangle to sit on top of the thing already constructed, you would tend to look back and forth.",
+      "output": "If your intention is to move the red triangle to sit on top of the thing already constructed, you would tend to look back and forth.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 6,
+      "category": "prose",
+      "transcript": "Um one thing that would be convenient for all sorts of things and which I haven't found out about it is how to do tool tips in an arbitrary way in in in swing graphic displays.",
+      "target": "One thing that would be convenient for all sorts of things, and which I haven't found out about, is how to do tooltips in an arbitrary way in Swing graphic displays.",
+      "output": "One thing that would be convenient for all sorts of things, and which I haven't found out about, is how to do tool tips in an arbitrary way in in Swing graphics displays.",
+      "exact": false,
+      "similarity": 0.985,
+      "target_coverage": 0.871,
+      "forbidden": []
+    },
+    {
+      "i": 7,
+      "category": "prose",
+      "transcript": "What we've finally just uh at twelve o'clock today, looks like we've got it sussed, um, is to get one channel one microphone being the left channel and one being the right channel to merge into a stereo single file.",
+      "target": "What we've finally just at twelve o'clock today, looks like we've got it sussed, is to get one channel, one microphone being the left channel, and one being the right channel, to merge into a stereo single file.",
+      "output": "At 12 o'clock today, looks like we've got it sussed, um, is to get one channel1 and one microphone.",
+      "exact": false,
+      "similarity": 0.155,
+      "target_coverage": 0.375,
+      "forbidden": []
+    },
+    {
+      "i": 8,
+      "category": "prose",
+      "transcript": "I'm just thinking of you know, ill utterance consisting entirely of laughter and then putting like punctuation in like ha ha ha ha just",
+      "target": "I'm just thinking of, you know, an utterance consisting entirely of laughter, and then putting like punctuation in, like...",
+      "output": "I'm just thinking of you.",
+      "exact": false,
+      "similarity": 0.338,
+      "target_coverage": 0.263,
+      "forbidden": []
+    },
+    {
+      "i": 9,
+      "category": "prose",
+      "transcript": "So we could zip we could zip through this and look at all the um all the interactions in which people built subcontracts first, which we expect them to do as a wild hypothesis more when they have verbal communication than when they don't.",
+      "target": "So we could zip through this and look at all the interactions in which people build subconstructs first, which we expect them to do, but as a wild hypothesis more when they have verbal communication than when they don't.",
+      "output": "We could zip through this and look at all the interactions in which people built subcontracts. First, they should do as a wild hypothesis, more when they have verbal communication than when they don't.",
+      "exact": false,
+      "similarity": 0.375,
+      "target_coverage": 0.692,
+      "forbidden": []
+    },
+    {
+      "i": 10,
+      "category": "prose",
+      "transcript": "I'm I I'm getting quite lost um at the moment because um what's um our difference between the um s um uh the importance measure and the skimming?",
+      "target": "I'm getting quite lost at the moment because what's our difference between the importance measure and the skimming?",
+      "output": "Our difference between the importance measure and the skimming.",
+      "exact": false,
+      "similarity": 0.685,
+      "target_coverage": 0.421,
+      "forbidden": []
+    },
+    {
+      "i": 11,
+      "category": "prose",
+      "transcript": "Which if you just said that it's gone Peterson absorption from neutral hydrogen rather than i if it was a if it was hydrogen absorption",
+      "target": "Which, if you just said that it's Gunn-Petersen absorption from neutral hydrogen rather than\u2014if it was a\u2014if it was hydrogen absorption.",
+      "output": "- If you just said that it's gone, Peterson absorption from neutral hydrogen rather than I\n- If it was a,",
+      "exact": false,
+      "similarity": 0.75,
+      "target_coverage": 0.667,
+      "forbidden": []
+    },
+    {
+      "i": 12,
+      "category": "prose",
+      "transcript": "So what I'm having is so we we had a we had a long problem about not finding out why the time syncing up between the relative time in a in a sort of in a in a summarized section and the absolute time didn't correspond and stuff.",
+      "target": "So what I'm having is, so we had a long problem about not finding out why the time-syncing up between the relative time in a sort of summarised section and the absolute time didn't correspond and stuff.",
+      "output": "So what I'm having is so we had a long problem about not finding out why the time syncing up between the relative time in a in a sort of in a summarized section and the absolute time didn't correspond.",
+      "exact": false,
+      "similarity": 0.864,
+      "target_coverage": 0.897,
+      "forbidden": []
+    },
+    {
+      "i": 13,
+      "category": "meeting",
+      "transcript": "So interactivity is a concept is a concept that I'm using to distinguish the two uh uh extremes of uh of interact uh of discourse in meetings.",
+      "target": "So 'interactivity' is a concept that I'm using to distinguish the two extremes of discourse in meetings.",
+      "output": "Interactivity is a concept, is a concept that I'm using to distinguish the two.",
+      "exact": false,
+      "similarity": 0.699,
+      "target_coverage": 0.647,
+      "forbidden": []
+    },
+    {
+      "i": 14,
+      "category": "intent",
+      "transcript": "And um so far um I think uh it's uh I think it it's uh still um about the s about the same level of segmentation we are talking here.",
+      "target": "And so far I think it's still about the same level of segmentation we're talking here.",
+      "output": "Still about the same level of segmentation we are talking here.",
+      "exact": false,
+      "similarity": 0.805,
+      "target_coverage": 0.471,
+      "forbidden": []
+    },
+    {
+      "i": 15,
+      "category": "prose",
+      "transcript": "As long as other people don't for some reason, I don't know, disconnect some bits of the equipment, there's no reason why they should.",
+      "target": "That as long as other people don't, for some reason\u2014I don't know, disconnect some bits of the equipment\u2014there's no reason why they should.",
+      "output": "As long as other people don't for some reason, I don't know, disconnect some bits of the equipment, there's no reason why they should.",
+      "exact": false,
+      "similarity": 0.949,
+      "target_coverage": 0.963,
+      "forbidden": []
+    },
+    {
+      "i": 16,
+      "category": "prose",
+      "transcript": "It's fairly general. Um and I guess the purpose of this is also not to restrict you in in the way you're thinking, like um, voice recognition.",
+      "target": "It's fairly general, and I guess the purpose of this is also not to restrict you in the way you're thinking, like voice recognition.",
+      "output": "It's fairly general. And I guess the purpose of this is also not to restrict you in the way you're thinking, like:\n\n- Voice recognition",
+      "exact": false,
+      "similarity": 0.959,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 17,
+      "category": "prose",
+      "transcript": "W within the sentence that you're sort of currently looking at, um feel free to correct it that way that you like it's not only about resolution but sort of like coming up with a corrected version on these on these places.",
+      "target": "Within the sentence that you sort of currently looking at, feel free to correct it that way, that you'll do, you know, like, it's not only about resolution but it's sort of like coming up with a corrected version based on these.",
+      "output": "Sort of currently looking at, feel free to correct it that way, you like it's not only about resolution but coming up with a corrected version on these places.",
+      "exact": false,
+      "similarity": 0.682,
+      "target_coverage": 0.591,
+      "forbidden": []
+    },
+    {
+      "i": 18,
+      "category": "prose",
+      "transcript": "And there will also be presumably some kind of task hierarchy modeling going on in I mean,'cause the whole browser working group um final um sort of presentation on that second day was about",
+      "target": "And there will also be, presumably, some kind of task hierarchy modelling going on, in\u2014I mean, 'cause the whole browser working group final, sort of, presentation on that second day was about...",
+      "output": "And there will also be presumably some kind of task hierarchy modeling going on in IME.",
+      "exact": false,
+      "similarity": 0.598,
+      "target_coverage": 0.375,
+      "forbidden": []
+    },
+    {
+      "i": 19,
+      "category": "prose",
+      "transcript": "Uh right, well basically um high priority for any animal for me is that they be willing to take a lot of physical affection from their family.",
+      "target": "Right, well, basically high priority for any animal, for me, is that they be willing to take a lot of physical affection from their family.",
+      "output": "High-priority for any animal for me is that they be willing to take a lot of physical affection from their family.",
+      "exact": false,
+      "similarity": 0.885,
+      "target_coverage": 0.875,
+      "forbidden": []
+    },
+    {
+      "i": 20,
+      "category": "prose",
+      "transcript": "It took like the third time till I thought, Oh, you're saying F five.",
+      "target": "It took, like, the third time till I thought, oh, you're saying F-five.",
+      "output": "It took like the third time till I thought, Oh, you're saying F5.",
+      "exact": false,
+      "similarity": 0.926,
+      "target_coverage": 0.857,
+      "forbidden": []
+    },
+    {
+      "i": 21,
+      "category": "prose",
+      "transcript": "Because, you know, algorithmically you're you're already putting together I mean this is essentially already, you know, these fixations with stuff in between.",
+      "target": "Because, you know, algorithmically, you're already putting together\u2014I mean, this is essentially already, you know, these fixations with stuff in between.",
+      "output": "These fixations with stuff in between.",
+      "exact": false,
+      "similarity": 0.387,
+      "target_coverage": 0.227,
+      "forbidden": []
+    },
+    {
+      "i": 22,
+      "category": "prose",
+      "transcript": "Yeah. Uh right, damp life enough. See again, that's uh these are source types. Um I mean it's not a source type.",
+      "target": "Right, damped Lyman\u2014see, again, that's\u2014these are source types\u2014the, I mean, it's not a source type.",
+      "output": "These are source types.",
+      "exact": false,
+      "similarity": 0.364,
+      "target_coverage": 0.158,
+      "forbidden": []
+    },
+    {
+      "i": 23,
+      "category": "prose",
+      "transcript": "So mm so for dialogue, yeah, mm there would be uh some kind of uh interaction structure.",
+      "target": "So for dialogue, yeah, there would be some kind of interaction structure.",
+      "output": "...",
+      "exact": false,
+      "similarity": 0.026,
+      "target_coverage": 0.0,
+      "forbidden": []
+    },
+    {
+      "i": 24,
+      "category": "prose",
+      "transcript": "Well, I could look'cause I I've got that there's the list, isn't there, of postgraduates that I maintain in in linguistics.",
+      "target": "Well, I could look, 'cause I've got\u2014that's the list, isn't there, of postgraduates I maintain in linguistics.",
+      "output": "I've got that, there's the list, isn't there, of postgraduates that I maintain in linguistics.",
+      "exact": false,
+      "similarity": 0.798,
+      "target_coverage": 0.65,
+      "forbidden": [
+        "here's the"
+      ]
+    },
+    {
+      "i": 25,
+      "category": "prose",
+      "transcript": "Um but the extra step that we're asking you to do in this case is to look at the silent regions and listen to them for each channel and um make sure that no back channels were lost, for example.",
+      "target": "But the extra step that we're asking you to do in this case is to look at the silent regions and listen to them for each channel, and make sure that no back channels were lost, for example.",
+      "output": "But the extra step that we're asking you to do in this case is to look at the silent regions and listen to them for each channel.",
+      "exact": false,
+      "similarity": 0.811,
+      "target_coverage": 0.711,
+      "forbidden": []
+    },
+    {
+      "i": 26,
+      "category": "meeting",
+      "transcript": "Are they doing um uh are are they uh as in the kickoff meeting just exchanging information to get started with the project or are they uh actually addressing some specific design um criterion at the moment?",
+      "target": "Are they doing, are they\u2014as in the kick-off meeting just exchanging information to get started with a project, or are they actually addressing some specific design criteria at the moment?",
+      "output": "Are they doing...?",
+      "exact": false,
+      "similarity": 0.146,
+      "target_coverage": 0.129,
+      "forbidden": []
+    },
+    {
+      "i": 27,
+      "category": "prose",
+      "transcript": "Would we wanna ask them and put some kind of like proviso in to make it um specific to sort of preemptively",
+      "target": "Would we wanna ask them and put some kind of, like, provisor in to make it specific, to sort of preemptively...",
+      "output": "Would we wanna ask them and put some kind of proviso in to make it specific to sort of preemptively?",
+      "exact": false,
+      "similarity": 0.938,
+      "target_coverage": 0.85,
+      "forbidden": []
+    },
+    {
+      "i": 28,
+      "category": "prose",
+      "transcript": "And there's especially one there that um I think he's Eastern European and I think that's the only reason I can understand what he's saying at all.",
+      "target": "And there's especially one there that, I think he's Eastern European, and I think that's the only reason I can understand what he's saying at all.",
+      "output": "And there's especially one there that I think he's Eastern European and I think that's the only reason I can understand what he's saying at all.",
+      "exact": false,
+      "similarity": 0.993,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 29,
+      "category": "meeting",
+      "transcript": "Because um I I think it's actually very similar to what I did for my um speaker um uh extraction and I think I would perhaps have to change two lines of code to get you um for each meeting a file that says fr from um this millisecond to this millisecond there was this sequence of words.",
+      "target": "Because I think it's actually very similar to what I did for my speaker extraction, and I think I would perhaps have to change two lines of code to get you, for each meeting, a file that says from this millisecond to this millisecond there was this sequence of words.",
+      "output": "Because I think it's actually very similar to what I did for my speaker, extraction and I think I would perhaps have to change two lines of code to get you for each meeting. a file that says FR from um this millisecond to this millisecond.",
+      "exact": false,
+      "similarity": 0.64,
+      "target_coverage": 0.84,
+      "forbidden": []
+    },
+    {
+      "i": 30,
+      "category": "prose",
+      "transcript": "I would just try to keep segments very manageable. Um and but y you use these uh waveforms to help you.",
+      "target": "I would just try to keep segments very manageable, but, you know, use these waveforms to help you.",
+      "output": "Use these waveforms to help you.",
+      "exact": false,
+      "similarity": 0.477,
+      "target_coverage": 0.294,
+      "forbidden": []
+    },
+    {
+      "i": 31,
+      "category": "prose",
+      "transcript": "Um right into the call that we are considering the idea of having a poster session.",
+      "target": "Write into the call that we are considering the idea of having a poster session.",
+      "output": "Right into the call that we are considering the idea of having a poster session.",
+      "exact": false,
+      "similarity": 0.963,
+      "target_coverage": 0.929,
+      "forbidden": []
+    },
+    {
+      "i": 32,
+      "category": "prose",
+      "transcript": "Well I'm I'm I'm I'm trying tr I'm making sure that the GDA format we've got at the moment will will output the well output all the analysis that that Robin's looking for.",
+      "target": "Well, I'm just trying\u2014I'm making sure that the GDF format we've got at the moment will output\u2014say, well, I've output old analysis that Robin's looking for.",
+      "output": "Trying tr, making sure that the GDA format we've got at the moment will output the well output, all the analysis that Robin's looking for.",
+      "exact": false,
+      "similarity": 0.826,
+      "target_coverage": 0.5,
+      "forbidden": []
+    },
+    {
+      "i": 33,
+      "category": "prose",
+      "transcript": "I can just add SSH stuff in to ev evoke the sound tools on the uh on dice and then copy them back.",
+      "target": "I can just add SSH stuff in to evoke the sound tools on Dice, and then copy 'em back.",
+      "output": "Add SSH stuff in to evoke the sound tools on the UH, and then copy them back.",
+      "exact": false,
+      "similarity": 0.852,
+      "target_coverage": 0.611,
+      "forbidden": []
+    },
+    {
+      "i": 34,
+      "category": "prose",
+      "transcript": "have all the l lines of actually lots of different elements and you have to work out like all of the ones which are iron lines.",
+      "target": "You have all the lines of actually lots of different elements, and you have to work out at all the ones which are ion lines.",
+      "output": "Have all the L lines of actually lots of different elements and you have to work out like all of the ones which are iron lines.",
+      "exact": false,
+      "similarity": 0.924,
+      "target_coverage": 0.792,
+      "forbidden": []
+    },
+    {
+      "i": 35,
+      "category": "prose",
+      "transcript": "Um suppose it doesn't really matter as long as they're unique and it can I can pass them back to you and you can use that to the display.",
+      "target": "I suppose it doesn't really matter as long as they're unique, and I can pass 'em back to you and you can use that to the display.",
+      "output": "Suppose it doesn't really matter, as long as they're unique and it can.",
+      "exact": false,
+      "similarity": 0.67,
+      "target_coverage": 0.429,
+      "forbidden": []
+    },
+    {
+      "i": 36,
+      "category": "prose",
+      "transcript": "So um I don't know, just just to to ask now if you've got anything else you've thought about while we've been talking.",
+      "target": "So, dunno, just to ask now if you've got anything else you've thought about while we've been talking.",
+      "output": "Just to ask now, if you've got anything else you've thought about while we've been talking.",
+      "exact": false,
+      "similarity": 0.927,
+      "target_coverage": 0.9,
+      "forbidden": []
+    },
+    {
+      "i": 37,
+      "category": "meeting",
+      "transcript": "I mean there probably are a bunch of issues like if you've got a hundred hours of meetings and they all talk about a lot of the same things.",
+      "target": "I mean, there probably are a bunch of issues, like, if you've got a hundred hours of meetings and they all talk about a lot of the same things.",
+      "output": "There probably are a bunch of issues.",
+      "exact": false,
+      "similarity": 0.4,
+      "target_coverage": 0.207,
+      "forbidden": []
+    },
+    {
+      "i": 38,
+      "category": "prose",
+      "transcript": "Um at least it works if there are several categories with each with one document each, but it um",
+      "target": "At least it works if there are several categories, with each, with one document each, but it\u2014yeah.",
+      "output": "At least it works if there are several categories, with each one document.",
+      "exact": false,
+      "similarity": 0.86,
+      "target_coverage": 0.647,
+      "forbidden": []
+    },
+    {
+      "i": 39,
+      "category": "prose",
+      "transcript": "Suppose we pick up an we we we screwed up the first time we put the triangle with the square and we threw it away. We decided we didn't like it.",
+      "target": "Suppose we screwed up the first time we put the triangle with the square, and we threw it away and decided we didn't like it.",
+      "output": "We decided we didn't like it.",
+      "exact": false,
+      "similarity": 0.364,
+      "target_coverage": 0.2,
+      "forbidden": []
+    },
+    {
+      "i": 40,
+      "category": "prose",
+      "transcript": "I think proportional like I even though I'm I'm always getting myself confused about this, but I think sort of making the summary length thing generally proportional",
+      "target": "I think proportion, like, even though I'm always getting myself confused about this, but I think sort of making the summary length thing generally proportion.",
+      "output": "I think proportional, but I'm always getting myself confused about this, but I'm making the summary length thing generally proportional.",
+      "exact": false,
+      "similarity": 0.871,
+      "target_coverage": 0.64,
+      "forbidden": []
+    },
+    {
+      "i": 41,
+      "category": "prose",
+      "transcript": "Uh what el uh well I'll I'll have to get to grips with Cra Camtasia uh properly. Uh to generate a little sort of introductory video, I suppose, uh for instructional purposes at the beginning of the experiment.",
+      "target": "Well, I'll have to get to grips with Camtasia properly to generate a little sort of introductory video, I suppose, for instructional purposes at the beginning of the experiment.",
+      "output": "Uh...well, I'll have to get to grips with CraCamtasia, properly.",
+      "exact": false,
+      "similarity": 0.448,
+      "target_coverage": 0.276,
+      "forbidden": []
+    },
+    {
+      "i": 42,
+      "category": "prose",
+      "transcript": "Um what I because nine megabyte is what I got for when I said for every um utterance. This goes from there to there and",
+      "target": "What I\u2014because nine megabytes is what I got for when I said for every utterance, this goes from there to there, and...",
+      "output": "This goes from there to there.",
+      "exact": false,
+      "similarity": 0.392,
+      "target_coverage": 0.227,
+      "forbidden": []
+    },
+    {
+      "i": 43,
+      "category": "prose",
+      "transcript": "If via the wasn't in there then you'd have it silicon two lines at but without with via the I think that means silicon two derived from these two features.",
+      "target": "If it wasn't in there, then you'd have it 'silicone two lines at' but without\u2014with the\u2014I think that means silicone two derived from these two features.",
+      "output": "If via the wasn't in there, you'd have it Silicon2.",
+      "exact": false,
+      "similarity": 0.436,
+      "target_coverage": 0.207,
+      "forbidden": []
+    },
+    {
+      "i": 44,
+      "category": "prose",
+      "transcript": "So what I've been thinking is maybe if we try to really make sense together of the XML format to be sure that we can all produce the data that we are producing in a way or at least I mean, I guess my data I'll probably load in from completely different way anyway because it's a matrix.",
+      "target": "So what I've been thinking is, maybe if we try to really make sense together of the XML format to be sure that we can all produce the data that we are producing in a way\u2014or at least, I mean, I guess my data I'll probably load in from completely different way anyway because it's matrix.",
+      "output": "So what I've been thinking is maybe if we try to really make sense together of the XML format, to be sure that we can all produce the data that we are producing in a way, or at least I'll remember, I'll probably load in from completely different way anyway because it's a matrix.",
+      "exact": false,
+      "similarity": 0.396,
+      "target_coverage": 0.881,
+      "forbidden": []
+    },
+    {
+      "i": 45,
+      "category": "prose",
+      "transcript": "Mm. Cause uh so for the information and action the uh the action here is not direct but uh some kind of uh uh physical action uh that uh the uh participants uh i is going to uh or has done in the past.",
+      "target": "'Cause, so, for the information and action, the action here is not dialogue act but some kind of physical action that the participant is going to or has done in the past.",
+      "output": "...",
+      "exact": false,
+      "similarity": 0.012,
+      "target_coverage": 0.0,
+      "forbidden": []
+    },
+    {
+      "i": 46,
+      "category": "meeting",
+      "transcript": "J because uh after talking to Jim, um just uh we have different ideas whether um meeting action will be about the same size as topic segments.",
+      "target": "Because after talking to Jane, just\u2014we have different ideas whether meeting action will be about the same size as topic segments.",
+      "output": "J because uh after talking to Jim,\n\n- meeting\n- action\n- will\n- be\n- about\n- topic\n- segments",
+      "exact": false,
+      "similarity": 0.631,
+      "target_coverage": 0.333,
+      "forbidden": []
+    },
+    {
+      "i": 47,
+      "category": "prose",
+      "transcript": "uh i it's also going to be a problem, uh, because you're going to have one part or construct forming in the middle of a fixation.",
+      "target": "It's also going to be a problem because you're going to have one part or a construct forming in the middle of a fixation.",
+      "output": "It's also going to be a problem, because you're going to have one part or constructing forming in the middle of a fixation.",
+      "exact": false,
+      "similarity": 0.975,
+      "target_coverage": 0.88,
+      "forbidden": []
+    },
+    {
+      "i": 48,
+      "category": "prose",
+      "transcript": "Um it ties you down perhaps a bit later on in terms of the technologies, um, how far you can transmit a signal or the complexity of the functions that you want.",
+      "target": "It ties you down perhaps a bit later on in terms of the technologies, how far you can transmit the signal, or the complexity of the functions that you want.",
+      "output": "It ties you down, perhaps a bit later on in terms of the technologies.",
+      "exact": false,
+      "similarity": 0.611,
+      "target_coverage": 0.483,
+      "forbidden": []
+    },
+    {
+      "i": 49,
+      "category": "prose",
+      "transcript": "Like that initially had quite different ideas about the phrase types and and markations and then eventually you came up with some kind of rules that you set up between uh the two of you.",
+      "target": "Like, that initially had quite different ideas about the phrase types and markations, and then eventually you came up with some kind of rules that you set up between the two of you.",
+      "output": "Initially had quite different ideas about the phrase types and markings. Then eventually you came up with some kind of rules that you set up between.",
+      "exact": false,
+      "similarity": 0.879,
+      "target_coverage": 0.719,
+      "forbidden": []
+    },
+    {
+      "i": 50,
+      "category": "meeting",
+      "transcript": "And so I think going through the presentations that um were given at the work package six meeting where um I think it was um it wasn't w was it Anita or was it oh I've forgotten her name.",
+      "target": "And so I think going through the presentations that were given at the work package six meeting, where, and I think it was\u2014it wasn't\u2014was it Anita, or was it\u2014oh, I've forgotten her name.",
+      "output": "And so I think going through the presentations that were given at the work package 6 meeting, where I think it was it wasn't w was it.",
+      "exact": false,
+      "similarity": 0.824,
+      "target_coverage": 0.622,
+      "forbidden": []
+    },
+    {
+      "i": 51,
+      "category": "prose",
+      "transcript": "Um frequencies or something. Um as well as um characters, um different uh keypad styles and symbols.",
+      "target": "Frequencies or something, as well as characters, different keypad styles and symbols.",
+      "output": "Frequencies:\n- Keypad styles\n- Symbols",
+      "exact": false,
+      "similarity": 0.504,
+      "target_coverage": 0.091,
+      "forbidden": []
+    },
+    {
+      "i": 52,
+      "category": "prose",
+      "transcript": "a couple of hours getting photocopies of degree certificates to give you for your files if I'm just gonna be here'cause they want I think it that one place it said give us copies of degree",
+      "target": "A couple of hours getting photocopies of degree certificates to give you for your files, if I'm just gonna be here a good\u2014they won't, I think at one place it said give us copies of degrees.",
+      "output": "A couple of hours getting photocopies of degree certificates to give you for your files. If I'm just gonna be here, because they want. It's give us copies of degree.",
+      "exact": false,
+      "similarity": 0.864,
+      "target_coverage": 0.632,
+      "forbidden": []
+    },
+    {
+      "i": 53,
+      "category": "prose",
+      "transcript": "What um the kinds of events that we had uh before talked about putting into the record for then use in Alan and NXT isn't based on saying every X second something",
+      "target": "What\u2014the kinds of events that we had before talked about putting into the record for then use in ELAN and NXT isn't based on saying every X seconds something.",
+      "output": "The kinds of events that we had before talked about putting into the record for then use in Alan and NXT isn't based on saying every X second.",
+      "exact": false,
+      "similarity": 0.92,
+      "target_coverage": 0.833,
+      "forbidden": []
+    },
+    {
+      "i": 54,
+      "category": "prose",
+      "transcript": "Mm we were doing descriptions that made things separate classes, in which case I'd argue for elliptical, I think.",
+      "target": "We were doing descriptions that made things separate classes, in which case I'd argue for elliptical, I think.",
+      "output": "We were doing descriptions that made things separate classes, in which case I'd argue for elliptical, I think.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 55,
+      "category": "meeting",
+      "transcript": "So uh in meetings I uh from my uh reading of the one of the M meetings IS one oh eight uh the uh the four meetings",
+      "target": "So in meetings, from my reading of one of the AMI meetings, IS one O eight, the four meetings.",
+      "output": "In meetings:\n- Reading of the one of the M meetings\n- The four meetings",
+      "exact": false,
+      "similarity": 0.715,
+      "target_coverage": 0.333,
+      "forbidden": []
+    },
+    {
+      "i": 56,
+      "category": "meeting",
+      "transcript": "And then between us we'll have to have another meeting and and I mean, at the same time as we're sorting through all the other titles and and deciding um how to divide up the sessions.",
+      "target": "And then between us we'll have to have another meeting, and I mean, at the same time as we're sorting through all the other titles and deciding how to divide up the sessions.",
+      "output": "And then between us we'll have to have another meeting.",
+      "exact": false,
+      "similarity": 0.48,
+      "target_coverage": 0.294,
+      "forbidden": []
+    },
+    {
+      "i": 57,
+      "category": "prose",
+      "transcript": "So this person left out a lot of N punctuation and um here's a case where I think she was transcribing an outbreath, um, or possibly an in breath, I uh in and used the hash mark.",
+      "target": "So this person left out a lot of end punctuation, and here's a case where I think she was transcribing an out-breath, or possibly an in-breath, in, and used the hash mark.",
+      "output": "Here's a case where I think she was transcribing an outbreath, I use the hash mark.",
+      "exact": false,
+      "similarity": 0.638,
+      "target_coverage": 0.353,
+      "forbidden": []
+    },
+    {
+      "i": 58,
+      "category": "prose",
+      "transcript": "And I actually uh think that that's a slightly extreme um position um with regard to the literature that a lot of people might say it doesn't really matter who the different speakers are in an interactive section.",
+      "target": "And I actually think that that's a slightly extreme position with regard to the literature, that a lot of people might say it doesn't really matter who the different speakers are in an interactive section.",
+      "output": "And I actually think that's a slightly extreme position, with regard to the literature that a lot of people might say it doesn't really matter.",
+      "exact": false,
+      "similarity": 0.345,
+      "target_coverage": 0.722,
+      "forbidden": []
+    },
+    {
+      "i": 59,
+      "category": "prose",
+      "transcript": "Oh, I had one guy, one of the project managers was um he had a really weird accent and he was calling um one guy Mr. Red.",
+      "target": "Oh, I had one guy\u2014one of the project managers was, he had a really weird accent and he was calling one guy 'Mister Ed'.",
+      "output": "Oh, I had one guy, one of the project managers was he had a really weird accent and he was calling one guy Mr. Red.",
+      "exact": false,
+      "similarity": 0.923,
+      "target_coverage": 0.917,
+      "forbidden": []
+    },
+    {
+      "i": 60,
+      "category": "prose",
+      "transcript": "By now I just had a look at the year data. Um and actually the average F zero deviation is only slightly higher for the positive ones than for the negative ones.",
+      "target": "By now, I just had a look at the data, and actually the average F-zero deviation is only slightly higher for the positive ones than for the negative ones.",
+      "output": "By now, I just had a look at the year data. And actually, the average F0 deviation is only slightly higher for the positive ones than for the negative ones.",
+      "exact": false,
+      "similarity": 0.948,
+      "target_coverage": 0.862,
+      "forbidden": []
+    },
+    {
+      "i": 61,
+      "category": "prose",
+      "transcript": "Um I would recommend I've found that it's easier when you're checking and you're trying to make sure the boundaries are right for a speech segment.",
+      "target": "I would recommend\u2014I've found that it's easier when you're checking and you're trying to make sure the boundaries are right for a speech segment.",
+      "output": "I would recommend I've found that it's easier when you're checking and you're trying to make sure the boundaries are right for a speech segment.",
+      "exact": false,
+      "similarity": 0.993,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 62,
+      "category": "prose",
+      "transcript": "But the the text for the actual call, is she going to rework something or uh I think she should",
+      "target": "But the text for the actual call, is she going to rework something, or she should?",
+      "output": "She should rework something.",
+      "exact": false,
+      "similarity": 0.418,
+      "target_coverage": 0.133,
+      "forbidden": []
+    },
+    {
+      "i": 63,
+      "category": "prose",
+      "transcript": "We've got we've got all the data in the ASCII format at the moment, and it's just a matter of what data do we need to pull out of that into the GDF so I can actually do the prop analysis.",
+      "target": "'Cause we got all this in the ASCII format at the moment, and it's just a matter of what data do we need to pull out of that into the GDF, so I can actually do the proper analysis.",
+      "output": "We've got all the data in the ASCII format at the moment, and it's just a matter of what data we need to pull out of that into the GDF so I can actually do the prop analysis.",
+      "exact": false,
+      "similarity": 0.932,
+      "target_coverage": 0.795,
+      "forbidden": []
+    },
+    {
+      "i": 64,
+      "category": "prose",
+      "transcript": "I think it's'cause we had to specify it ourselves that it's not as um like focus a specification of most um work we have to do.",
+      "target": "I think it's 'cause we had to specify it ourselves, that it's not as, like, focus the specification of most work we have to do.",
+      "output": "We had to do.",
+      "exact": false,
+      "similarity": 0.171,
+      "target_coverage": 0.115,
+      "forbidden": []
+    },
+    {
+      "i": 65,
+      "category": "prose",
+      "transcript": "Yeah, Dave C in the program manager, do you have a a method to like set the current ID so it'll update the display?",
+      "target": "Yeah, Dave, see, in the programme manager, do you have a method to, like, set the current ID so it'll update the display?",
+      "output": "Yeah, Dave C in the program manager, do you have a method to set the current ID so it'll update the display?",
+      "exact": false,
+      "similarity": 0.934,
+      "target_coverage": 0.739,
+      "forbidden": []
+    },
+    {
+      "i": 66,
+      "category": "prose",
+      "transcript": "um which is the the first one is that um uh the companies decided that teletext is outdated uh because of how popular the internet is.",
+      "target": "Which is the first one, is that the company's decided that teletext is outdated because of how popular the internet is.",
+      "output": "The companies decided that teletext is outdated, because of how popular the internet is.",
+      "exact": false,
+      "similarity": 0.812,
+      "target_coverage": 0.524,
+      "forbidden": []
+    },
+    {
+      "i": 67,
+      "category": "meeting",
+      "transcript": "So should we expect that all the slides from like the work package six meeting or the joint work package five work pack will be up on one of the work package websites?",
+      "target": "So should we expect that all the slides from, like, the work package six meeting or the joint work package five work package will be up on one of the work package websites?",
+      "output": "Should we expect that all the slides from the work package 6 meeting will be up on one of the work package websites?",
+      "exact": false,
+      "similarity": 0.799,
+      "target_coverage": 0.625,
+      "forbidden": []
+    },
+    {
+      "i": 68,
+      "category": "prose",
+      "transcript": "If we need just frequencies, maybe we should just calculate them by using Pearl or something.",
+      "target": "If we need just frequencies, maybe we should just calculate them by using Perl or something.",
+      "output": "If we need just frequencies, maybe we should just calculate them by using Pearl.",
+      "exact": false,
+      "similarity": 0.919,
+      "target_coverage": 0.8,
+      "forbidden": []
+    },
+    {
+      "i": 69,
+      "category": "prose",
+      "transcript": "So so essentially there's this um A E is divided into T R one and C and other stuff. Okay, where C is the construct, the existing thing.",
+      "target": "So essentially there's a AE, is divided into TR-one and C, and other stuff, okay, where C is the construct, the existing thing.",
+      "output": "Essentially, there's this: A, E, is divided into T, R, and C, and other stuff.",
+      "exact": false,
+      "similarity": 0.644,
+      "target_coverage": 0.333,
+      "forbidden": []
+    },
+    {
+      "i": 70,
+      "category": "prose",
+      "transcript": "If I have to do it f like flexibly then I probably then it'll not be significantly more difficult.",
+      "target": "If I have to do it flexibly, then I probably\u2014then it will not be significantly more difficult.",
+      "output": "If I have to do it, flexibly then I'll not be significantly more difficult.",
+      "exact": false,
+      "similarity": 0.864,
+      "target_coverage": 0.706,
+      "forbidden": []
+    },
+    {
+      "i": 71,
+      "category": "prose",
+      "transcript": "And I've uh sent a thing off to Marlo's talk trying to encourage her again to register for the C S V and uh and what to C V S and uh wh what to do and uh which bit of software I use that seems to be quite easy. I was just imagining that okay, and you never",
+      "target": "And I've sent a thing off to Marloes Talk trying to encourage her again to register for the CVS, and what to\u2014the CVS, and what to do, and which bit of software I use, that seems to be quite easy and simple.",
+      "output": "And I've sent a thing off to Marlo's talk, trying to encourage her again to register for the C S V and...",
+      "exact": false,
+      "similarity": 0.592,
+      "target_coverage": 0.372,
+      "forbidden": []
+    },
+    {
+      "i": 72,
+      "category": "prose",
+      "transcript": "Yeah, what I did at the moment is um I um yeah, I got the F zero values from for each speaker for his headphone.",
+      "target": "Yeah, what I did at the moment is I got the F-zero values from\u2014for each speaker, for his headphone.",
+      "output": "Yeah, what I did at the moment is:\n\n- I got the F0 values from for each speaker for his headphone.",
+      "exact": false,
+      "similarity": 0.934,
+      "target_coverage": 0.85,
+      "forbidden": []
+    },
+    {
+      "i": 73,
+      "category": "prose",
+      "transcript": "Oh gosh, narrow emission Okay, so I've highlighted this in red and then the whole thing in green because type two C foot galaxies are one.",
+      "target": "Oh gosh, no emission line, okay, so I've highlighted this in red, and then the whole thing in green because type two Seyfart galaxies are one.",
+      "output": "Oh, gosh, narrow emission \u2014OK, so I've highlighted this in red and then the whole thing in green because type 2 C-food galaxies are one.",
+      "exact": false,
+      "similarity": 0.878,
+      "target_coverage": 0.692,
+      "forbidden": []
+    },
+    {
+      "i": 74,
+      "category": "meeting",
+      "transcript": "It more and more appears to me that if we if we scrap the notion of the meeting as an individual thing and sort of ch see meetings as as topic segments and have sort of like hierarchical topic segmentation instead.",
+      "target": "It more and more appears to me that if we scrap the notion of the meeting as an individual thing and sort of just see meetings as topic segments and have sort of, like, hierarchical topic segmentation instead.",
+      "output": "More and more appears to me that if we scrap the notion of the meeting as an individual thing and sort of ch see meetings as topic segments and have sort of like hierarchical topic segmentation instead.",
+      "exact": false,
+      "similarity": 0.968,
+      "target_coverage": 0.919,
+      "forbidden": []
+    },
+    {
+      "i": 75,
+      "category": "prose",
+      "transcript": "So from uh um my analysis of that kind of dialogues, I could say that there's only there are only four basic uh exchanges depend on the information uh direction of information flow.",
+      "target": "So from my analysis of that kind of dialogues, I could say that there are only four basic exchanges, depending on the direction of the information flow.",
+      "output": "From up to:\n\n- My analysis of that kind of dialogues, I could say that there's only there are only four basic exchanges.",
+      "exact": false,
+      "similarity": 0.699,
+      "target_coverage": 0.615,
+      "forbidden": []
+    },
+    {
+      "i": 76,
+      "category": "meeting",
+      "transcript": "I think they are more talking about um user requirement for the meeting than subtext they want to know.",
+      "target": "I think they are more talking about user requirement for the meeting than soft text they want to know.",
+      "output": "I think they are more talking about the user requirement for the meeting than subtext they want to know.",
+      "exact": false,
+      "similarity": 0.951,
+      "target_coverage": 0.778,
+      "forbidden": []
+    },
+    {
+      "i": 77,
+      "category": "prose",
+      "transcript": "Um I've sorted out uh travel, etcetera, to Munich for uh for Ellen, John and myself.",
+      "target": "I've sorted out travel et cetera to Munich for Ellen, John, and myself.",
+      "output": "I've sorted out...",
+      "exact": false,
+      "similarity": 0.36,
+      "target_coverage": 0.231,
+      "forbidden": []
+    },
+    {
+      "i": 78,
+      "category": "prose",
+      "transcript": "It does make sense from maybe the design point of view'cause if you have more complicated characters like European languages then you need more buttons.",
+      "target": "It does make sense from maybe the design point of view, 'cause you have more complicated characters, like European languages, then you need more buttons.",
+      "output": "It makes sense from a maybe the design point of view, because if you have more complicated characters like European languages then you need more buttons.",
+      "exact": false,
+      "similarity": 0.948,
+      "target_coverage": 0.75,
+      "forbidden": []
+    },
+    {
+      "i": 79,
+      "category": "prose",
+      "transcript": "So it's more like a it's uh it's sort of like a quite literal task. So it's really like finding these kind of uh these types of strings uh rather than really referring to the underlying physical reality.",
+      "target": "So it's more like\u2014it's sort of like a quite literal task that's really like finding these kind of, these types of strings, rather than really referring to the underlying physical reality.",
+      "output": "Sort of like a, it's quite literal.",
+      "exact": false,
+      "similarity": 0.261,
+      "target_coverage": 0.176,
+      "forbidden": []
+    },
+    {
+      "i": 80,
+      "category": "prose",
+      "transcript": "And yeah, that they have lots of personality and uh be fit and in robust good health.",
+      "target": "And yeah, that they have lots of personality, and be fit, and in robust good health.",
+      "output": "And yeah, that they have lots of personality and be fit and in robust good health.",
+      "exact": false,
+      "similarity": 0.988,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 81,
+      "category": "prose",
+      "transcript": "And sometimes with the ds uh if you're thinking, are you anticipating a d or you're you trying to say th and you're just saying the the the the the",
+      "target": "And sometimes with the\u2014I would be thinking, are you anticipating, or are you trying to say 'the' and you're just saying 'the the the the the'?",
+      "output": "...",
+      "exact": false,
+      "similarity": 0.0,
+      "target_coverage": 0.0,
+      "forbidden": []
+    },
+    {
+      "i": 82,
+      "category": "prose",
+      "transcript": "So I mean I I do think you should keep in mind what the final goal is so that you can plug this into something at the end and and um show some benefit of having the markup.",
+      "target": "So I mean, I do think you should keep in mind what the final goal is, so that you can plug this into something at the end and show some benefit of having the markup.",
+      "output": "I do think you should keep in mind what the final goal is, so that you can plug this into something at the end and show some benefit of having the markup.",
+      "exact": false,
+      "similarity": 0.966,
+      "target_coverage": 0.912,
+      "forbidden": []
+    },
+    {
+      "i": 83,
+      "category": "prose",
+      "transcript": "But you wouldn't be able to point to something in a spectrum and say that's a damped Lyman Alpha Absorber unless you'd",
+      "target": "But you wouldn't be able to point something in a spectrum and say that's a damped Lyman alpha absorber unless you'd...",
+      "output": "But you wouldn't be able to point to something in a spectrum and say that's a damped Lyman-Alpha Absorber unless you'd.",
+      "exact": false,
+      "similarity": 0.954,
+      "target_coverage": 0.957,
+      "forbidden": []
+    },
+    {
+      "i": 84,
+      "category": "meeting",
+      "transcript": "So uh and currently I'm more interested in the uh uh in analyzing the disconstruction multi party meetings.",
+      "target": "So, currently I'm more interested in, yeah, analysing the discourse structure in multi-party meetings.",
+      "output": "Currently I'm more interested in the disconstruction multi-party meetings.",
+      "exact": false,
+      "similarity": 0.807,
+      "target_coverage": 0.467,
+      "forbidden": []
+    },
+    {
+      "i": 85,
+      "category": "prose",
+      "transcript": "In fact, even by inviting um people from um informatics and from well maybe from philosophy and things, it might bias it in a particular direction.",
+      "target": "In fact, even by inviting people from Informatics and from\u2014or maybe from Philosophy and things\u2014it might bias it in a particular direction.",
+      "output": "In fact, even by inviting people from informatics and from philosophy and things, it might bias it in a particular direction.",
+      "exact": false,
+      "similarity": 0.92,
+      "target_coverage": 0.87,
+      "forbidden": []
+    },
+    {
+      "i": 86,
+      "category": "prose",
+      "transcript": "But they're they're creating a there. Uh guy, Jonathan, is creating a web form for you to do this all very simply and like just um upload everything that you've uh uh upload the file that you've transcribed and and you'll be noting whether it's the first pass or the second pass and your name um as the person who did that particular pass.",
+      "target": "But they're creating\u2014their guy Jonathan is creating a web form for you to do this all very simply, and like just upload everything that you've\u2014upload the file that you've transcribed, and you'll be noting whether it's the first pass or the second pass, and your name as the person who did that particular pass.",
+      "output": "Uh, guy, Jonathan, is creating a web form for you to do this, all very simply and like just Upload everything that you've.",
+      "exact": false,
+      "similarity": 0.042,
+      "target_coverage": 0.4,
+      "forbidden": []
+    },
+    {
+      "i": 87,
+      "category": "prose",
+      "transcript": "So one speaker will have the floored channel um throughout that. And so I guess what you're saying is that the interactive sections would have shorter speaker turns when you're not considering the back channels as interruptions.",
+      "target": "So one speaker will have the floor channel throughout that, and so I guess what you're saying is that the interactive sections would have shorter speaker turns when you're not considering the back channels as interruptions.",
+      "output": "...",
+      "exact": false,
+      "similarity": 0.009,
+      "target_coverage": 0.0,
+      "forbidden": []
+    },
+    {
+      "i": 88,
+      "category": "prose",
+      "transcript": "It's usually a context problem because I think sometimes if it's not intelligible, you either tend to because you really want it to fit the context",
+      "target": "It's usually a context problem, because I think sometimes if it's not intelligible, you either tend to\u2014because you really want it to fit the context.",
+      "output": "It's usually a context problem because I think sometimes if it's not intelligible, you either tend to because you really want it to fit the context.",
+      "exact": false,
+      "similarity": 0.99,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 89,
+      "category": "meeting",
+      "transcript": "Couldn't we couldn't we write one program that calculates all the um f for all our different contributions to the importance measure the um absolute um dialogue act numbers for the meeting?",
+      "target": "Couldn't we write one programme that calculates all the\u2014for all our different contributions to the importance measure, the absolute dialogue act numbers for the meeting?",
+      "output": "Couldn't we?\n\n-1",
+      "exact": false,
+      "similarity": 0.13,
+      "target_coverage": 0.077,
+      "forbidden": []
+    },
+    {
+      "i": 90,
+      "category": "prose",
+      "transcript": "So no um n no transcriber will be checking no checker will be checking their own transcription.",
+      "target": "So no transcriber will be checking\u2014no checker will be checking their own transcription.",
+      "output": "Transcriber will be checking their own transcription.",
+      "exact": false,
+      "similarity": 0.743,
+      "target_coverage": 0.615,
+      "forbidden": []
+    },
+    {
+      "i": 91,
+      "category": "prose",
+      "transcript": "In some cases we might if it's really tough, we might be in need of a of some sound system to play recordings.",
+      "target": "In some cases we might\u2014if it's really tough we might be in need of some sound system to play recordings.",
+      "output": "In some cases, we might be in need of a sound system to play recordings.",
+      "exact": false,
+      "similarity": 0.795,
+      "target_coverage": 0.619,
+      "forbidden": []
+    },
+    {
+      "i": 92,
+      "category": "prose",
+      "transcript": "So at the moment I'm looking at act actually the GDF format itself. Um and then to generate the GDA format I'm gonna be taking stuff out of the Jest analyzer um and then using the DOM parser to generate uh XML output from the Gest analyzer.",
+      "target": "So at the moment I'm looking at actually the GDF format itself, and then to generate the GDF format I'm gonna be taking stuff out of the gest analyser, and then using a DOM parser to generate XML output from the gest analyser.",
+      "output": "...",
+      "exact": false,
+      "similarity": 0.009,
+      "target_coverage": 0.0,
+      "forbidden": []
+    },
+    {
+      "i": 93,
+      "category": "prose",
+      "transcript": "Well what I think I might try and build is basically a class that you just feed it a linked list of um different waveforms.",
+      "target": "Well, what I think I might try and build is basically a class that you just feed it a linked list of different waveforms.",
+      "output": "Well, what I think I might try and build is basically a class that you just feed it a linked list of different waveforms.",
+      "exact": true,
+      "similarity": 1.0,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 94,
+      "category": "intent",
+      "transcript": "So I guess um if I'm gonna be segmenting it with a L C Seg then that's the same format I'd want to um put it back out in, so it'd be equivalent.",
+      "target": "So I guess if I'm gonna be segmenting it with an LCSeg, then that's, like, same format I'd want to put it back out in, so it'd be equivalent.",
+      "output": "...",
+      "exact": false,
+      "similarity": 0.014,
+      "target_coverage": 0.0,
+      "forbidden": []
+    },
+    {
+      "i": 95,
+      "category": "prose",
+      "transcript": "Well well a worry that was was expressed in the new requirements was that if we made this too complex it would it it would affect um how long how long it took us to get this to market.",
+      "target": "Well, a worry that was expressed in the new requirements was that, if we made this too complex, it would affect how long it took us to get this to market.",
+      "output": "Well, a worry that was expressed in the new requirements. Was that if we made this too complex it would, it would affect, how long, how long it took us to get this to market.",
+      "exact": false,
+      "similarity": 0.927,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 96,
+      "category": "prose",
+      "transcript": "Um and so um it is important that these relations have a semantics that's tied to the semantics of the utterances, right?",
+      "target": "And so it is important that these relations have a semantics that's tied to the semantics of the utterances, right?",
+      "output": "And so, it is important that these relations have a semantics that's tied to the semantics of the utterances, right?",
+      "exact": false,
+      "similarity": 0.996,
+      "target_coverage": 1.0,
+      "forbidden": []
+    },
+    {
+      "i": 97,
+      "category": "prose",
+      "transcript": "Yeah, that's what I thought as well. That you that probably the the topic segment level is the most um informative for the words.",
+      "target": "Yeah, that's what I thought as well, that probably the topic segment level is the most informative for the words.",
+      "output": "Yeah, that's what I thought, well. The topic segment level is the most informative for the words.",
+      "exact": false,
+      "similarity": 0.895,
+      "target_coverage": 0.75,
+      "forbidden": []
+    },
+    {
+      "i": 98,
+      "category": "prose",
+      "transcript": "If people are busy interacting all the time and all of our colleagues don't even bother to control for whether people are talking to one another when they're doing these joint tasks because it seems to them to be irrelevant because language is irrelevant.",
+      "target": "If people are busy interacting all the time and all of our colleagues don't even bother to control for whether people are talking to one another when they're doing these joint tasks, because it seems to them to be irrelevant, 'cause language is irrelevant.",
+      "output": "If people are busy interacting all the time and all of our colleagues don't even bother to control for whether people are talking to one another when they're doing these joint tasks, because it seems to them be irrelevant.",
+      "exact": false,
+      "similarity": 0.929,
+      "target_coverage": 0.867,
+      "forbidden": []
+    },
+    {
+      "i": 99,
+      "category": "prose",
+      "transcript": "It's sort of it's overrunning the f like it. So okay, so at the moment we are that far that if I now click onto a segment here it's",
+      "target": "It's sort of\u2014it's overrunning the F, like, it\u2014so okay, so at the moment we're that far that if I now click onto a segment, it hits.",
+      "output": "It's overrunning the f, like it.\n\nOkay, so at the moment we are that far. If I now click onto a segment.",
+      "exact": false,
+      "similarity": 0.809,
+      "target_coverage": 0.633,
+      "forbidden": []
+    }
+  ]
+}

lfm2_5_350m_config.json

@@ -0,0 +1,33 @@
+{
+  "dim": 1024,
+  "ffn_dim_multiplier": 1,
+  "hidden_dim": 4608,
+  "n_heads": 16,
+  "n_kv_heads": 8,
+  "n_layers": 16,
+  "norm_eps": 1e-5,
+  "rope_theta": 1000000.0,
+  "use_scaled_rope": false,
+  "vocab_size": 65536,
+  "use_hf_rope": true,
+  "use_qk_norm": true,
+  "qk_norm_before_rope": true,
+  "layer_types": [
+    "conv",
+    "conv",
+    "full_attention",
+    "conv",
+    "conv",
+    "full_attention",
+    "conv",
+    "conv",
+    "full_attention",
+    "conv",
+    "full_attention",
+    "conv",
+    "full_attention",
+    "conv",
+    "full_attention",
+    "conv"
+  ]
+}

lfm2_5_350m_ft.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa46b25ba5541e1e4bb6eaa1953235234f8fcbe1c978b5cad23ae91933b1c5fb
+size 1417985357

lfm2_5_350m_mlx_4w.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3e220c62890f3ac655e23068f2ec59aa2cdabeb74aa86a95a725d7d463bb853
+size 490342656

tokenizer/chat_template.jinja

@@ -0,0 +1,64 @@
+{{- bos_token -}}
+{%- set keep_past_thinking = keep_past_thinking | default(false) -%}
+{%- set ns = namespace(system_prompt="") -%}
+{%- if messages[0]["role"] == "system" -%}
+    {%- set sys_content = messages[0]["content"] -%}
+    {%- if sys_content is not string -%}
+        {%- for item in sys_content -%}
+            {%- if item["type"] == "text" -%}
+                {%- set ns.system_prompt = ns.system_prompt + item["text"] -%}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {%- set ns.system_prompt = sys_content -%}
+    {%- endif -%}
+    {%- set messages = messages[1:] -%}
+{%- endif -%}
+{%- if tools -%}
+    {%- set ns.system_prompt = ns.system_prompt + ("\n" if ns.system_prompt else "") + "List of tools: [" -%}
+    {%- for tool in tools -%}
+        {%- if tool is not string -%}
+            {%- set tool = tool | tojson -%}
+        {%- endif -%}
+        {%- set ns.system_prompt = ns.system_prompt + tool -%}
+        {%- if not loop.last -%}
+            {%- set ns.system_prompt = ns.system_prompt + ", " -%}
+        {%- endif -%}
+    {%- endfor -%}
+    {%- set ns.system_prompt = ns.system_prompt + "]" -%}
+{%- endif -%}
+{%- if ns.system_prompt -%}
+    {{- "<|im_start|>system\n" + ns.system_prompt + "<|im_end|>\n" -}}
+{%- endif -%}
+{%- set ns.last_assistant_index = -1 -%}
+{%- for message in messages -%}
+    {%- if message["role"] == "assistant" -%}
+        {%- set ns.last_assistant_index = loop.index0 -%}
+    {%- endif -%}
+{%- endfor -%}
+{%- for message in messages -%}
+    {{- "<|im_start|>" + message["role"] + "\n" -}}
+    {%- set content = message["content"] -%}
+    {%- if content is not string -%}
+        {%- set ns.content = "" -%}
+        {%- for item in content -%}
+            {%- if item["type"] == "image" -%}
+                {%- set ns.content = ns.content + "<image>" -%}
+            {%- elif item["type"] == "text" -%}
+                {%- set ns.content = ns.content + item["text"] -%}
+            {%- else -%}
+                {%- set ns.content = ns.content + item | tojson -%}
+            {%- endif -%}
+        {%- endfor -%}
+        {%- set content = ns.content -%}
+    {%- endif -%}
+    {%- if message["role"] == "assistant" and not keep_past_thinking and loop.index0 != ns.last_assistant_index -%}
+        {%- if "</think>" in content -%}
+            {%- set content = content.split("</think>")[-1] | trim -%}
+        {%- endif -%}
+    {%- endif -%}
+    {{- content + "<|im_end|>\n" -}}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{- "<|im_start|>assistant\n" -}}
+{%- endif -%}

tokenizer/special_tokens_map.json

@@ -0,0 +1,23 @@
+{
+  "bos_token": {
+    "content": "<|startoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|pad|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer/tokenizer_config.json

@@ -0,0 +1,4091 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<|startoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<|fim_pre|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<|fim_mid|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<|fim_suf|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<|tool_list_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<|tool_list_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<|tool_call_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<|tool_call_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<|tool_response_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<|tool_response_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<|reserved_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "17": {
+      "content": "<|reserved_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "18": {
+      "content": "<|reserved_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "19": {
+      "content": "<|reserved_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "20": {
+      "content": "<|reserved_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "21": {
+      "content": "<|reserved_11|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "22": {
+      "content": "<|reserved_12|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "23": {
+      "content": "<|reserved_13|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "24": {
+      "content": "<|reserved_14|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "25": {
+      "content": "<|reserved_15|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "26": {
+      "content": "<|reserved_16|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "27": {
+      "content": "<|reserved_17|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "28": {
+      "content": "<|reserved_18|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "29": {
+      "content": "<|reserved_19|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "30": {
+      "content": "<|reserved_20|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "31": {
+      "content": "<|reserved_21|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32": {
+      "content": "<|reserved_22|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "33": {
+      "content": "<|reserved_23|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "34": {
+      "content": "<|reserved_24|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "35": {
+      "content": "<|reserved_25|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "36": {
+      "content": "<|reserved_26|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "37": {
+      "content": "<|reserved_27|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "38": {
+      "content": "<|reserved_28|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "39": {
+      "content": "<|reserved_29|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "40": {
+      "content": "<|reserved_30|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "41": {
+      "content": "<|reserved_31|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "42": {
+      "content": "<|reserved_32|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "43": {
+      "content": "<|reserved_33|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "44": {
+      "content": "<|reserved_34|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "45": {
+      "content": "<|reserved_35|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "46": {
+      "content": "<|reserved_36|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "47": {
+      "content": "<|reserved_37|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "48": {
+      "content": "<|reserved_38|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "49": {
+      "content": "<|reserved_39|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "50": {
+      "content": "<|reserved_40|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "51": {
+      "content": "<|reserved_41|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "52": {
+      "content": "<|reserved_42|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "53": {
+      "content": "<|reserved_43|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "54": {
+      "content": "<|reserved_44|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "55": {
+      "content": "<|reserved_45|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "56": {
+      "content": "<|reserved_46|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "57": {
+      "content": "<|reserved_47|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "58": {
+      "content": "<|reserved_48|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "59": {
+      "content": "<|reserved_49|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "60": {
+      "content": "<|reserved_50|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "61": {
+      "content": "<|reserved_51|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "62": {
+      "content": "<|reserved_52|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "63": {
+      "content": "<|reserved_53|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64": {
+      "content": "<|reserved_54|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "65": {
+      "content": "<|reserved_55|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "66": {
+      "content": "<|reserved_56|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "67": {
+      "content": "<|reserved_57|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "68": {
+      "content": "<|reserved_58|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "69": {
+      "content": "<|reserved_59|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "70": {
+      "content": "<|reserved_60|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "71": {
+      "content": "<|reserved_61|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "72": {
+      "content": "<|reserved_62|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "73": {
+      "content": "<|reserved_63|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "74": {
+      "content": "<|reserved_64|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "75": {
+      "content": "<|reserved_65|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "76": {
+      "content": "<|reserved_66|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "77": {
+      "content": "<|reserved_67|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "78": {
+      "content": "<|reserved_68|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "79": {
+      "content": "<|reserved_69|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "80": {
+      "content": "<|reserved_70|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "81": {
+      "content": "<|reserved_71|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "82": {
+      "content": "<|reserved_72|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "83": {
+      "content": "<|reserved_73|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "84": {
+      "content": "<|reserved_74|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "85": {
+      "content": "<|reserved_75|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "86": {
+      "content": "<|reserved_76|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "87": {
+      "content": "<|reserved_77|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "88": {
+      "content": "<|reserved_78|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "89": {
+      "content": "<|reserved_79|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "90": {
+      "content": "<|reserved_80|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "91": {
+      "content": "<|reserved_81|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "92": {
+      "content": "<|reserved_82|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "93": {
+      "content": "<|reserved_83|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "94": {
+      "content": "<|reserved_84|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "95": {
+      "content": "<|reserved_85|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "96": {
+      "content": "<|reserved_86|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "97": {
+      "content": "<|reserved_87|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "98": {
+      "content": "<|reserved_88|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "99": {
+      "content": "<|reserved_89|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100": {
+      "content": "<|reserved_90|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "101": {
+      "content": "<|reserved_91|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "102": {
+      "content": "<|reserved_92|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "103": {
+      "content": "<|reserved_93|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "104": {
+      "content": "<|reserved_94|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "105": {
+      "content": "<|reserved_95|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "106": {
+      "content": "<|reserved_96|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "107": {
+      "content": "<|reserved_97|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "108": {
+      "content": "<|reserved_98|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "109": {
+      "content": "<|reserved_99|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "110": {
+      "content": "<|reserved_100|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "111": {
+      "content": "<|reserved_101|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "112": {
+      "content": "<|reserved_102|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "113": {
+      "content": "<|reserved_103|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "114": {
+      "content": "<|reserved_104|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "115": {
+      "content": "<|reserved_105|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "116": {
+      "content": "<|reserved_106|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "117": {
+      "content": "<|reserved_107|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "118": {
+      "content": "<|reserved_108|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "119": {
+      "content": "<|reserved_109|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "120": {
+      "content": "<|reserved_110|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "121": {
+      "content": "<|reserved_111|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "122": {
+      "content": "<|reserved_112|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "123": {
+      "content": "<|reserved_113|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "124": {
+      "content": "<|reserved_114|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "125": {
+      "content": "<|reserved_115|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "126": {
+      "content": "<|reserved_116|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "127": {
+      "content": "<|reserved_117|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "128": {
+      "content": "<|audio_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "129": {
+      "content": "<|text_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "130": {
+      "content": "<|text_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "131": {
+      "content": "<|mixed_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "132": {
+      "content": "<|mixed_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "133": {
+      "content": "<|reserved_123|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "134": {
+      "content": "<|reserved_124|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "135": {
+      "content": "<|reserved_125|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "136": {
+      "content": "<|reserved_126|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "137": {
+      "content": "<|reserved_127|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "138": {
+      "content": "<|reserved_128|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "139": {
+      "content": "<|reserved_129|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "140": {
+      "content": "<|reserved_130|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "141": {
+      "content": "<|reserved_131|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "142": {
+      "content": "<|reserved_132|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "143": {
+      "content": "<|reserved_133|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "144": {
+      "content": "<|reserved_134|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "145": {
+      "content": "<|reserved_135|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "146": {
+      "content": "<|reserved_136|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "147": {
+      "content": "<|reserved_137|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "148": {
+      "content": "<|reserved_138|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "149": {
+      "content": "<|reserved_139|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "150": {
+      "content": "<|reserved_140|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151": {
+      "content": "<|reserved_141|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "152": {
+      "content": "<|reserved_142|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "153": {
+      "content": "<|reserved_143|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "154": {
+      "content": "<|reserved_144|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "155": {
+      "content": "<|reserved_145|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "156": {
+      "content": "<|reserved_146|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "157": {
+      "content": "<|reserved_147|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "158": {
+      "content": "<|reserved_148|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "159": {
+      "content": "<|reserved_149|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "160": {
+      "content": "<|reserved_150|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "161": {
+      "content": "<|reserved_151|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "162": {
+      "content": "<|reserved_152|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "163": {
+      "content": "<|reserved_153|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "164": {
+      "content": "<|reserved_154|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "165": {
+      "content": "<|reserved_155|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "166": {
+      "content": "<|reserved_156|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "167": {
+      "content": "<|reserved_157|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "168": {
+      "content": "<|reserved_158|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "169": {
+      "content": "<|reserved_159|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "170": {
+      "content": "<|reserved_160|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "171": {
+      "content": "<|reserved_161|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "172": {
+      "content": "<|reserved_162|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "173": {
+      "content": "<|reserved_163|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "174": {
+      "content": "<|reserved_164|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "175": {
+      "content": "<|reserved_165|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "176": {
+      "content": "<|reserved_166|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "177": {
+      "content": "<|reserved_167|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "178": {
+      "content": "<|reserved_168|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "179": {
+      "content": "<|reserved_169|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "180": {
+      "content": "<|reserved_170|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "181": {
+      "content": "<|reserved_171|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "182": {
+      "content": "<|reserved_172|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "183": {
+      "content": "<|reserved_173|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "184": {
+      "content": "<|reserved_174|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "185": {
+      "content": "<|reserved_175|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "186": {
+      "content": "<|reserved_176|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "187": {
+      "content": "<|reserved_177|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "188": {
+      "content": "<|reserved_178|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "189": {
+      "content": "<|reserved_179|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "190": {
+      "content": "<|reserved_180|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "191": {
+      "content": "<|reserved_181|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "192": {
+      "content": "<|reserved_182|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "193": {
+      "content": "<|reserved_183|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "194": {
+      "content": "<|reserved_184|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "195": {
+      "content": "<|reserved_185|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "196": {
+      "content": "<|reserved_186|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "197": {
+      "content": "<|reserved_187|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "198": {
+      "content": "<|reserved_188|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "199": {
+      "content": "<|reserved_189|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "200": {
+      "content": "<|reserved_190|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "201": {
+      "content": "<|reserved_191|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "202": {
+      "content": "<|reserved_192|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "203": {
+      "content": "<|reserved_193|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "204": {
+      "content": "<|reserved_194|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "205": {
+      "content": "<|reserved_195|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "206": {
+      "content": "<|reserved_196|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "207": {
+      "content": "<|reserved_197|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "208": {
+      "content": "<|reserved_198|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "209": {
+      "content": "<|reserved_199|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "210": {
+      "content": "<|reserved_200|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "211": {
+      "content": "<|reserved_201|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "212": {
+      "content": "<|reserved_202|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "213": {
+      "content": "<|reserved_203|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "214": {
+      "content": "<|reserved_204|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "215": {
+      "content": "<|reserved_205|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "216": {
+      "content": "<|reserved_206|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "217": {
+      "content": "<|reserved_207|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "218": {
+      "content": "<|reserved_208|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "219": {
+      "content": "<|reserved_209|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "220": {
+      "content": "<|reserved_210|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "221": {
+      "content": "<|reserved_211|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "222": {
+      "content": "<|reserved_212|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "223": {
+      "content": "<|reserved_213|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "224": {
+      "content": "<|reserved_214|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "225": {
+      "content": "<|reserved_215|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "226": {
+      "content": "<|reserved_216|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "227": {
+      "content": "<|reserved_217|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "228": {
+      "content": "<|reserved_218|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "229": {
+      "content": "<|reserved_219|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "230": {
+      "content": "<|reserved_220|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "231": {
+      "content": "<|reserved_221|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "232": {
+      "content": "<|reserved_222|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "233": {
+      "content": "<|reserved_223|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "234": {
+      "content": "<|reserved_224|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "235": {
+      "content": "<|reserved_225|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "236": {
+      "content": "<|reserved_226|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "237": {
+      "content": "<|reserved_227|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "238": {
+      "content": "<|reserved_228|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "239": {
+      "content": "<|reserved_229|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "240": {
+      "content": "<|reserved_230|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "241": {
+      "content": "<|reserved_231|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "242": {
+      "content": "<|reserved_232|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "243": {
+      "content": "<|reserved_233|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "244": {
+      "content": "<|reserved_234|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "245": {
+      "content": "<|reserved_235|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "246": {
+      "content": "<|reserved_236|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "247": {
+      "content": "<|reserved_237|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "248": {
+      "content": "<|reserved_238|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "249": {
+      "content": "<|reserved_239|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "250": {
+      "content": "<|reserved_240|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "251": {
+      "content": "<|reserved_241|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "252": {
+      "content": "<|reserved_242|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "253": {
+      "content": "<|reserved_243|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "254": {
+      "content": "<|reserved_244|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "255": {
+      "content": "<|reserved_245|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "256": {
+      "content": "<|reserved_246|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "257": {
+      "content": "<|reserved_247|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "258": {
+      "content": "<|reserved_248|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "259": {
+      "content": "<|reserved_249|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "260": {
+      "content": "<|reserved_250|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "261": {
+      "content": "<|reserved_251|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "262": {
+      "content": "<|reserved_252|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "263": {
+      "content": "<|reserved_253|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "264": {
+      "content": "<|reserved_254|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "265": {
+      "content": "<|reserved_255|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "266": {
+      "content": "<|reserved_256|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "267": {
+      "content": "<|reserved_257|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "268": {
+      "content": "<|reserved_258|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "269": {
+      "content": "<|reserved_259|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "270": {
+      "content": "<|reserved_260|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "271": {
+      "content": "<|reserved_261|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "272": {
+      "content": "<|reserved_262|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "273": {
+      "content": "<|reserved_263|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "274": {
+      "content": "<|reserved_264|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "275": {
+      "content": "<|reserved_265|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "276": {
+      "content": "<|reserved_266|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "277": {
+      "content": "<|reserved_267|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "278": {
+      "content": "<|reserved_268|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "279": {
+      "content": "<|reserved_269|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "280": {
+      "content": "<|reserved_270|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "281": {
+      "content": "<|reserved_271|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "282": {
+      "content": "<|reserved_272|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "283": {
+      "content": "<|reserved_273|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "284": {
+      "content": "<|reserved_274|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "285": {
+      "content": "<|reserved_275|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "286": {
+      "content": "<|reserved_276|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "287": {
+      "content": "<|reserved_277|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "288": {
+      "content": "<|reserved_278|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "289": {
+      "content": "<|reserved_279|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "290": {
+      "content": "<|reserved_280|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "291": {
+      "content": "<|reserved_281|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "292": {
+      "content": "<|reserved_282|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "293": {
+      "content": "<|reserved_283|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "294": {
+      "content": "<|reserved_284|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "295": {
+      "content": "<|reserved_285|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "296": {
+      "content": "<|reserved_286|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "297": {
+      "content": "<|reserved_287|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "298": {
+      "content": "<|reserved_288|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "299": {
+      "content": "<|reserved_289|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "300": {
+      "content": "<|reserved_290|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "301": {
+      "content": "<|reserved_291|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "302": {
+      "content": "<|reserved_292|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "303": {
+      "content": "<|reserved_293|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "304": {
+      "content": "<|reserved_294|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "305": {
+      "content": "<|reserved_295|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "306": {
+      "content": "<|reserved_296|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "307": {
+      "content": "<|reserved_297|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "308": {
+      "content": "<|reserved_298|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "309": {
+      "content": "<|reserved_299|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "310": {
+      "content": "<|reserved_300|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "311": {
+      "content": "<|reserved_301|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "312": {
+      "content": "<|reserved_302|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "313": {
+      "content": "<|reserved_303|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "314": {
+      "content": "<|reserved_304|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "315": {
+      "content": "<|reserved_305|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "316": {
+      "content": "<|reserved_306|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "317": {
+      "content": "<|reserved_307|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "318": {
+      "content": "<|reserved_308|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "319": {
+      "content": "<|reserved_309|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "320": {
+      "content": "<|reserved_310|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "321": {
+      "content": "<|reserved_311|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "322": {
+      "content": "<|reserved_312|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "323": {
+      "content": "<|reserved_313|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "324": {
+      "content": "<|reserved_314|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "325": {
+      "content": "<|reserved_315|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "326": {
+      "content": "<|reserved_316|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "327": {
+      "content": "<|reserved_317|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "328": {
+      "content": "<|reserved_318|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "329": {
+      "content": "<|reserved_319|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "330": {
+      "content": "<|reserved_320|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "331": {
+      "content": "<|reserved_321|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "332": {
+      "content": "<|reserved_322|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "333": {
+      "content": "<|reserved_323|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "334": {
+      "content": "<|reserved_324|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "335": {
+      "content": "<|reserved_325|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "336": {
+      "content": "<|reserved_326|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "337": {
+      "content": "<|reserved_327|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "338": {
+      "content": "<|reserved_328|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "339": {
+      "content": "<|reserved_329|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "340": {
+      "content": "<|reserved_330|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "341": {
+      "content": "<|reserved_331|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "342": {
+      "content": "<|reserved_332|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "343": {
+      "content": "<|reserved_333|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "344": {
+      "content": "<|reserved_334|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "345": {
+      "content": "<|reserved_335|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "346": {
+      "content": "<|reserved_336|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "347": {
+      "content": "<|reserved_337|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "348": {
+      "content": "<|reserved_338|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "349": {
+      "content": "<|reserved_339|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "350": {
+      "content": "<|reserved_340|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "351": {
+      "content": "<|reserved_341|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "352": {
+      "content": "<|reserved_342|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "353": {
+      "content": "<|reserved_343|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "354": {
+      "content": "<|reserved_344|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "355": {
+      "content": "<|reserved_345|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "356": {
+      "content": "<|reserved_346|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "357": {
+      "content": "<|reserved_347|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "358": {
+      "content": "<|reserved_348|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "359": {
+      "content": "<|reserved_349|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "360": {
+      "content": "<|reserved_350|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "361": {
+      "content": "<|reserved_351|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "362": {
+      "content": "<|reserved_352|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "363": {
+      "content": "<|reserved_353|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "364": {
+      "content": "<|reserved_354|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "365": {
+      "content": "<|reserved_355|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "366": {
+      "content": "<|reserved_356|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "367": {
+      "content": "<|reserved_357|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "368": {
+      "content": "<|reserved_358|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "369": {
+      "content": "<|reserved_359|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "370": {
+      "content": "<|reserved_360|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "371": {
+      "content": "<|reserved_361|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "372": {
+      "content": "<|reserved_362|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "373": {
+      "content": "<|reserved_363|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "374": {
+      "content": "<|reserved_364|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "375": {
+      "content": "<|reserved_365|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "376": {
+      "content": "<|reserved_366|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "377": {
+      "content": "<|reserved_367|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "378": {
+      "content": "<|reserved_368|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "379": {
+      "content": "<|reserved_369|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "380": {
+      "content": "<|reserved_370|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "381": {
+      "content": "<|reserved_371|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "382": {
+      "content": "<|reserved_372|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "383": {
+      "content": "<|reserved_373|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "384": {
+      "content": "<|reserved_374|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "385": {
+      "content": "<|reserved_375|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "386": {
+      "content": "<|reserved_376|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "387": {
+      "content": "<|reserved_377|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "388": {
+      "content": "<|reserved_378|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "389": {
+      "content": "<|reserved_379|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "390": {
+      "content": "<|reserved_380|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "391": {
+      "content": "<|reserved_381|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "392": {
+      "content": "<|reserved_382|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "393": {
+      "content": "<|reserved_383|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "394": {
+      "content": "<|reserved_384|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "395": {
+      "content": "<|reserved_385|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "396": {
+      "content": "<image>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "397": {
+      "content": "<|img_row_1_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "398": {
+      "content": "<|img_row_1_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "399": {
+      "content": "<|img_row_1_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "400": {
+      "content": "<|img_row_1_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "401": {
+      "content": "<|img_row_1_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "402": {
+      "content": "<|img_row_1_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "403": {
+      "content": "<|img_row_1_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "404": {
+      "content": "<|img_row_1_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "405": {
+      "content": "<|img_row_1_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "406": {
+      "content": "<|img_row_1_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "407": {
+      "content": "<|img_row_2_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "408": {
+      "content": "<|img_row_2_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "409": {
+      "content": "<|img_row_2_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "410": {
+      "content": "<|img_row_2_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "411": {
+      "content": "<|img_row_2_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "412": {
+      "content": "<|img_row_2_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "413": {
+      "content": "<|img_row_2_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "414": {
+      "content": "<|img_row_2_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "415": {
+      "content": "<|img_row_2_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "416": {
+      "content": "<|img_row_2_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "417": {
+      "content": "<|img_row_3_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "418": {
+      "content": "<|img_row_3_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "419": {
+      "content": "<|img_row_3_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "420": {
+      "content": "<|img_row_3_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "421": {
+      "content": "<|img_row_3_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "422": {
+      "content": "<|img_row_3_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "423": {
+      "content": "<|img_row_3_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "424": {
+      "content": "<|img_row_3_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "425": {
+      "content": "<|img_row_3_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "426": {
+      "content": "<|img_row_3_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "427": {
+      "content": "<|img_row_4_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "428": {
+      "content": "<|img_row_4_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "429": {
+      "content": "<|img_row_4_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "430": {
+      "content": "<|img_row_4_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "431": {
+      "content": "<|img_row_4_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "432": {
+      "content": "<|img_row_4_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "433": {
+      "content": "<|img_row_4_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "434": {
+      "content": "<|img_row_4_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "435": {
+      "content": "<|img_row_4_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "436": {
+      "content": "<|img_row_4_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "437": {
+      "content": "<|img_row_5_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "438": {
+      "content": "<|img_row_5_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "439": {
+      "content": "<|img_row_5_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "440": {
+      "content": "<|img_row_5_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "441": {
+      "content": "<|img_row_5_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "442": {
+      "content": "<|img_row_5_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "443": {
+      "content": "<|img_row_5_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "444": {
+      "content": "<|img_row_5_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "445": {
+      "content": "<|img_row_5_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "446": {
+      "content": "<|img_row_5_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "447": {
+      "content": "<|img_row_6_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "448": {
+      "content": "<|img_row_6_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "449": {
+      "content": "<|img_row_6_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "450": {
+      "content": "<|img_row_6_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "451": {
+      "content": "<|img_row_6_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "452": {
+      "content": "<|img_row_6_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "453": {
+      "content": "<|img_row_6_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "454": {
+      "content": "<|img_row_6_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "455": {
+      "content": "<|img_row_6_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "456": {
+      "content": "<|img_row_6_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "457": {
+      "content": "<|img_row_7_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "458": {
+      "content": "<|img_row_7_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "459": {
+      "content": "<|img_row_7_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "460": {
+      "content": "<|img_row_7_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "461": {
+      "content": "<|img_row_7_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "462": {
+      "content": "<|img_row_7_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "463": {
+      "content": "<|img_row_7_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "464": {
+      "content": "<|img_row_7_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "465": {
+      "content": "<|img_row_7_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "466": {
+      "content": "<|img_row_7_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "467": {
+      "content": "<|img_row_8_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "468": {
+      "content": "<|img_row_8_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "469": {
+      "content": "<|img_row_8_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "470": {
+      "content": "<|img_row_8_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "471": {
+      "content": "<|img_row_8_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "472": {
+      "content": "<|img_row_8_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "473": {
+      "content": "<|img_row_8_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "474": {
+      "content": "<|img_row_8_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "475": {
+      "content": "<|img_row_8_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "476": {
+      "content": "<|img_row_8_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "477": {
+      "content": "<|img_row_9_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "478": {
+      "content": "<|img_row_9_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "479": {
+      "content": "<|img_row_9_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "480": {
+      "content": "<|img_row_9_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "481": {
+      "content": "<|img_row_9_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "482": {
+      "content": "<|img_row_9_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "483": {
+      "content": "<|img_row_9_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "484": {
+      "content": "<|img_row_9_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "485": {
+      "content": "<|img_row_9_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "486": {
+      "content": "<|img_row_9_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "487": {
+      "content": "<|img_row_10_col_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "488": {
+      "content": "<|img_row_10_col_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "489": {
+      "content": "<|img_row_10_col_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "490": {
+      "content": "<|img_row_10_col_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "491": {
+      "content": "<|img_row_10_col_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "492": {
+      "content": "<|img_row_10_col_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "493": {
+      "content": "<|img_row_10_col_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "494": {
+      "content": "<|img_row_10_col_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "495": {
+      "content": "<|img_row_10_col_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "496": {
+      "content": "<|img_row_10_col_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "497": {
+      "content": "<|img_thumbnail|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "498": {
+      "content": "<|image_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "499": {
+      "content": "<|image_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "500": {
+      "content": "<|image_split|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64011": {
+      "content": "Mathias",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64014": {
+      "content": "python",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64394": {
+      "content": "<|cot_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64395": {
+      "content": "<|cot_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64396": {
+      "content": "<|review_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64397": {
+      "content": "<|review_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64398": {
+      "content": "<|file_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64399": {
+      "content": "<|file_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64400": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64401": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|startoftext|>",
+  "chat_template_file": "chat_template.jinja",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "extra_special_tokens": {},
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 32768,
+  "pad_token": "<|pad|>",
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "use_default_system_prompt": false,
+  "use_fast": true
+}