Atomic Chat self-quantized LiquidAI/LFM2.5-8B-A1B ladder (imatrix, calibration_datav3)

.gitattributes

@@ -33,3 +33,15 @@ 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
+lfm25-8b-a1b-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-Q2_K.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+lfm25-8b-a1b-UD-Q4_K_XL.gguf filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,158 @@
+---
+license: other
+license_name: lfm1.0
+license_link: https://huggingface.co/LiquidAI/LFM2.5-8B-A1B/raw/main/LICENSE
+thumbnail: https://huggingface.co/AlexAtomic/lfm25-8b-a1b-GGUF/resolve/main/hero.png
+base_model:
+- LiquidAI/LFM2.5-8B-A1B
+base_model_relation: quantized
+quantized_by: AlexAtomic
+language:
+- en
+- ar
+- zh
+- fr
+- de
+- ja
+- ko
+- es
+- pt
+- it
+pipeline_tag: text-generation
+library_name: gguf
+tags:
+- atomic-chat
+- lfm
+- liquid
+- lfm2
+- gguf
+- imatrix
+- quantized
+- llama.cpp
+---
+
+<center>
+
+<div style="display:flex; justify-content:center; align-items:center; gap:10px; flex-wrap:wrap;">
+<a href="https://atomic.chat"><img src="https://huggingface.co/AlexAtomic/lfm25-8b-a1b-GGUF/resolve/main/pill_atomic_v3.png" alt="Atomic Chat" width="186"></a>
+<a href="https://discord.gg/8wGSsvmg4V"><img src="https://huggingface.co/AlexAtomic/lfm25-8b-a1b-GGUF/resolve/main/pill_discord_v3.png" alt="Join Discord" width="184"></a>
+<a href="https://github.com/AtomicBot-ai/Atomic-Chat"><img src="https://huggingface.co/AlexAtomic/lfm25-8b-a1b-GGUF/resolve/main/pill_github_v3.png" alt="GitHub" width="141"></a>
+</div>
+
+<br/>
+
+<img src="https://huggingface.co/AlexAtomic/lfm25-8b-a1b-GGUF/resolve/main/hero.png" alt="LFM2.5 8B A1B" style="width:100%; max-width:100%; height:auto; margin-bottom:0.6em;"/>
+
+<div style="display:flex; justify-content:center; gap:0.5em;">
+<a href="https://huggingface.co/LiquidAI/LFM2.5-8B-A1B"><strong>Base model: LiquidAI/LFM2.5-8B-A1B</strong></a>
+</div>
+</center>
+
+**LFM2.5 8B A1B**, self-quantized to GGUF by [Atomic Chat](https://atomic.chat). Built straight from Liquid AI's original weights with a per-tensor importance matrix. Runs fully offline.
+
+## Highlights
+
+- **Sparse MoE**: 8.3B total parameters, only 1.5B active per token.
+- **LFM2 hybrid architecture**: 24 layers (18 double-gated LIV convolution blocks + 6 GQA attention), built on LFM2 with extended pre-training and reinforcement learning.
+- **On-device assistant**: designed to chain tool calls and follow complex instructions, with day-one support for llama.cpp, MLX, vLLM and SGLang.
+- **Reasoning model**: assistant turns include an explicit chain of thought before the final answer.
+- **128K context**, 128,000 vocabulary, trained on a 38 trillion token budget.
+- **Multilingual**: English, Arabic, Chinese, French, German, Italian, Japanese, Korean, Portuguese, Spanish.
+
+> [!NOTE]
+> These GGUFs are **self-quantized from the original weights**, not a repack. The importance matrix keeps low-bit quants closer to the full-precision model.
+
+> [!IMPORTANT]
+> Always pass `--jinja` so the **LFM2.5 8B A1B chat template** is applied. Without it the model can emit malformed turns.
+
+## Model Overview
+
+| Property | Value |
+|---|---|
+| Base model | `LiquidAI/LFM2.5-8B-A1B` |
+| Total / active parameters | 8.3B total, 1.5B active (MoE) |
+| Layers | 24 (18 LIV conv + 6 GQA) |
+| Context length | 128,000 |
+| Architecture | LFM2.5 hybrid (built on LFM2, extended pre-training + RL) |
+| This repo | GGUF quants (imatrix) |
+
+<img src="https://huggingface.co/AlexAtomic/lfm25-8b-a1b-GGUF/resolve/main/benchmark.png" alt="LFM2.5 8B A1B benchmark scores" style="width:100%; max-width:900px;"/>
+
+Scores are Liquid AI's published results for the base `LiquidAI/LFM2.5-8B-A1B`. Quantization preserves the large majority of this; `Q4_K_M` and up sit within a point or two of full precision.
+
+## Quality
+
+KL-divergence and top-token agreement measured against our full-precision `f16` over 40 chunks. **Lower KLD = closer to full precision.**
+
+| Quant | Mean KLD ↓ | Top-token match ↑ |
+|---|---|---|
+| `Q4_K_M` — Atomic | 0.126876 | 82.896% |
+| `Q4_K_M` — Unsloth | 0.074829 | 87.029% |
+
+Unsloth comparison: [`unsloth/LFM2.5-8B-A1B-GGUF`](https://huggingface.co/unsloth/LFM2.5-8B-A1B-GGUF), identical test set and chunk count.
+
+## Choosing a quant
+
+| Quant | Size | Notes |
+|---|---|---|
+| `Q2_K` | 3.2 GB | Smallest. Minimal RAM, clear quality drop. |
+| `IQ3_M` | 3.8 GB | Beats Q3 at similar size thanks to imatrix. Best low-RAM pick. |
+| `Q3_K_M` | 4.1 GB | Low quality but usable. |
+| `Q3_K_L` | 4.4 GB | A step above Q3_K_M. |
+| `IQ4_XS` | 4.6 GB | Excellent quality for size. Recommended low-bit. |
+| `Q4_K_S` | 4.9 GB | Compact Q4, fast. |
+| **`Q4_K_M`** | 5.2 GB | **Recommended default. Best balance of size, speed and quality.** |
+| **`UD-Q4_K_XL`** | 5.2 GB | **Dynamic. Embeddings and output kept at Q8_0 for higher quality at a Q4 footprint.** |
+| `Q5_K_S` | 5.9 GB | Higher quality. |
+| `Q5_K_M` | 6.0 GB | Higher quality, low loss. |
+| `Q6_K` | 7.0 GB | Near lossless. |
+| `Q8_0` | 9.0 GB | Effectively lossless, reference quality. |
+
+> [!TIP]
+> Pick the largest file that fits your (V)RAM with room for context. `Q4_K_M` or `UD-Q4_K_XL` is the sweet spot for most setups; `Q6_K` or `Q8_0` for maximum fidelity.
+
+## Get started
+
+Run LFM2.5 8B A1B locally with:
+
+- **[Atomic Chat](https://atomic.chat):** the easiest path. Open the app, search `AlexAtomic/lfm25-8b-a1b-GGUF`, pick a quant, hit **Use this model**.
+- **llama.cpp:** `llama-server -hf AlexAtomic/lfm25-8b-a1b-GGUF:Q4_K_M --jinja -c 8192`
+- **Ollama:** `ollama run hf.co/AlexAtomic/lfm25-8b-a1b-GGUF:Q4_K_M`
+- **LM Studio / Jan:** search the repo id, download any quant.
+
+## Best practices
+
+| Parameter | Value |
+|---|---|
+| temperature | 0.2 |
+| top_k | 80 |
+| repetition_penalty | 1.05 |
+
+Liquid AI's recommended generation parameters.
+
+## Run in llama.cpp
+
+```bash
+git clone https://github.com/ggerganov/llama.cpp
+cmake llama.cpp -B llama.cpp/build -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=ON
+cmake --build llama.cpp/build --config Release -j --target llama-cli llama-server
+```
+
+```bash
+./llama.cpp/build/bin/llama-server \
+    -hf AlexAtomic/lfm25-8b-a1b-GGUF:UD-Q4_K_XL \
+    --jinja -ngl 99 -c 8192 -fa on
+```
+
+## How these were made
+
+1. Download `LiquidAI/LFM2.5-8B-A1B` (original weights).
+2. Convert to f16 GGUF with [llama.cpp](https://github.com/ggerganov/llama.cpp).
+3. Build an importance matrix over `calibration_datav3` (100 chunks).
+4. Quantize the full ladder with `--imatrix`.
+5. `UD-Q4_K_XL` additionally pins the token-embedding and output tensors to `Q8_0`.
+
+## License
+
+Released by Liquid AI under their LFM1.0 license. Quantized by Atomic Chat.
+

bench.json

@@ -0,0 +1,6 @@
+{
+  "model": "lfm25-8b-a1b", "chunks": 40, "f16_ppl": "36.0986", "unsloth_repo": "unsloth/LFM2.5-8B-A1B-GGUF",
+  "rows": [
+    {"quant":"Q4_K_M","our_kld":"0.126876","our_agree":"82.896","our_ppl":"2.792977","uns_kld":"0.074829","uns_agree":"87.029","uns_ppl":"0.225532"}
+  ]
+}

lfm25-8b-a1b-IQ3_M.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b007291d6ffa0b987966f85eeba2505ebd4f6d6717eb64f70d9bbac6a74f16e
+size 3778752032

lfm25-8b-a1b-IQ4_XS.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:025bbb71cf0d43e45a718fd92897233d820faa06b52b4ba52b7c22d1a2999a29
+size 4588301856

lfm25-8b-a1b-Q2_K.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bafc634954578c9e68aef058656b4fa7da639cc4baade0cdce6f6af588eb6f2
+size 3190435360

lfm25-8b-a1b-Q3_K_L.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed612ef1055216c3789144926d688b63dbae4b51bb9ad46010db955334e716c3
+size 4436864544

lfm25-8b-a1b-Q3_K_M.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:593b211efe50169e13f4ec81a0889956d9024e61bd3a162d7635928665d2bf38
+size 4108398112

lfm25-8b-a1b-Q4_K_M.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72955dd6c06a8833e660af0aa1cc9a4d722cdc92abe94a61715ae7b866d67bf2
+size 5155565088

lfm25-8b-a1b-Q4_K_S.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6feceb7b07b4252f72c158c2394eae3392304af805900841013934fd3ce5831
+size 4863553056

lfm25-8b-a1b-Q5_K_M.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d547400a4486771d93ca195204965b172ae2c0ca2dbd3344fa7f44b4b2c716c3
+size 6030339616

lfm25-8b-a1b-Q5_K_S.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:992df0d7b33873c3c79033de757b18f46ac7f0184c9c1ae92bbcd1ee1a46a757
+size 5870186016

lfm25-8b-a1b-Q6_K.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e9fbc921e3d7f93ed81db80cce917dbc8d142a1d28c365c95abe30cce9050b9
+size 6959787552

lfm25-8b-a1b-Q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67b3b20eaab2d0eb87eb971ceb1e0d7b229c8bc3b90cbb46d33f0ecfa5b17e63
+size 9010196000

lfm25-8b-a1b-UD-Q4_K_XL.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:283b12943743c5bb54b7f9fc8f7076c8ea32d163a06fb0e6f525178e7232c588
+size 5219053088