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
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2026 Alibaba Cloud
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

MERGE_COMPLETE

@@ -0,0 +1 @@
+2026-07-09T22:37:29.976640+00:00

README.md

@@ -0,0 +1,94 @@
+---
+library_name: transformers
+license: apache-2.0
+license_link: https://huggingface.co/Qwen/Qwen3.6-35B-A3B/blob/995ad96eacd98c81ed38be0c5b274b04031597b0/LICENSE
+base_model: Qwen/Qwen3.6-35B-A3B
+pipeline_tag: image-text-to-text
+tags:
+  - qwen3.6
+  - moe
+  - lora
+  - merged
+  - antidoom
+---
+
+# Qwen3.6-35B-A3B-AntiLoop
+
+This is a BF16 merged-weight variant of
+[`Qwen/Qwen3.6-35B-A3B`](https://huggingface.co/Qwen/Qwen3.6-35B-A3B),
+fine-tuned to recover from pathological self-verification and enumeration loops
+while preserving ordinary long-form reasoning.
+
+The artifact contains a full Transformers checkpoint; no PEFT adapter is needed
+at inference time. It retains the base model's multimodal architecture,
+tokenizer, chat template, MTP weights, and 262,144-token native context length.
+
+## Provenance
+
+- Base: `Qwen/Qwen3.6-35B-A3B`
+- Pinned base revision: `995ad96eacd98c81ed38be0c5b274b04031597b0`
+- Fine-tuning: rank-32 LoRA, alpha 64, dropout 0
+- Targets: 190 attention-side projections across all 40 language-model layers
+  (150 Gated DeltaNet projections and 40 full-attention projections)
+- Objective: `0.9 * supervised CE + 0.1 * KL(base || tuned)`
+- Training: one epoch / 178 optimizer steps, seed 0
+- Training run: `n8programs/qwen-antidoom`, run `dw7mbmce`
+
+The LoRA was merged as
+`BF16(W_FP32 + B_FP32 @ (A_FP32 * 2))`, one target tensor at a time in
+small row tiles. `merge_manifest.json` records the exact adapter hash, base
+revision, environment, per-target effective deltas, and every output-shard
+SHA-256. `merge_verification.json` certifies that all 190 intended tensors equal
+that formula bit-for-bit after BF16 rounding, every target differs from the
+base, and every non-target byte is unchanged.
+
+## Evaluation
+
+The behavior evaluations below used the same round-2 adapter served on the
+Qwen3.6-35B-A3B FP8 base with vLLM. They validate the adapter behavior; the
+merged BF16 artifact was separately validated at the tensor level.
+
+| Evaluation | Base/control | AntiLoop |
+|---|---:|---:|
+| Held-out enumeration-loop rate (285 prompts, calibrated judge) | 25% | **2%** |
+| Answer delivery on the same prompts | 77% | **98%** |
+| GPQA Diamond, thinking enabled | 167/198 (84.3%) | **166/198 (83.8%)** |
+
+The paired GPQA difference was -0.5 percentage points (exact two-sided McNemar
+`p = 1.0`), consistent with preserved capability at this sample size. The loop
+evaluation is task-specific and judge-based; it should not be read as a general
+safety or truthfulness guarantee.
+
+## Usage
+
+Use a recent Transformers release with Qwen3.6 support:
+
+```python
+from transformers import AutoModelForImageTextToText, AutoProcessor
+
+model_id = "N8Programs/Qwen3.6-35B-A3B-AntiLoop"
+processor = AutoProcessor.from_pretrained(model_id)
+model = AutoModelForImageTextToText.from_pretrained(
+    model_id,
+    dtype="auto",
+    device_map="auto",
+)
+```
+
+The checkpoint can also be served directly with recent vLLM/SGLang builds that
+support the Qwen3.6 MoE architecture. Follow the upstream Qwen3.6 model card for
+chat templating, thinking-mode controls, multimodal inputs, deployment details,
+and the base model's limitations.
+
+## Limitations
+
+- This is a narrow behavioral fine-tune, not a general alignment or safety model.
+- The merged BF16 checkpoint has not been exhaustively evaluated across every
+  language, modality, tool-use setting, or long-context regime.
+- Quantization can change behavior; validate the exact deployed quantization.
+- Outputs may still be incorrect, overconfident, repetitive, or unsafe.
+
+## License
+
+Apache 2.0, following the base checkpoint. See `LICENSE` and the upstream
+[Qwen3.6-35B-A3B model card](https://huggingface.co/Qwen/Qwen3.6-35B-A3B).

VERIFIED

@@ -0,0 +1 @@
+2026-07-09T22:43:31.178589+00:00

antiloop_training_args.json

@@ -0,0 +1,31 @@
+{
+  "model": "Qwen/Qwen3.6-35B-A3B",
+  "sft_file": "data/antidoom_sft_train_v2.jsonl",
+  "kl_file": "data/antidoom_kl_anchor.jsonl",
+  "output_dir": "checkpoints/qwen36_35b_antidoom_kl_lora_r32_v2",
+  "epochs": 1,
+  "max_steps": -1,
+  "max_seq_len": 6144,
+  "kl_max_seq_len": 4096,
+  "ce_weight": 0.9,
+  "kl_weight": 0.1,
+  "lr": 5e-05,
+  "warmup_ratio": 0.03,
+  "weight_decay": 0.0,
+  "grad_clip": 1.0,
+  "lora_r": 32,
+  "lora_alpha": 64,
+  "lora_dropout": 0.0,
+  "target_preset": "attn",
+  "seed": 0,
+  "loss_chunk": 4096,
+  "kl_chunk": 1024,
+  "attn_implementation": "sdpa",
+  "fp8_experts": true,
+  "fp8_model": "Qwen/Qwen3.6-35B-A3B-FP8",
+  "no_grad_checkpoint": true,
+  "log_every": 5,
+  "wandb_project": "qwen-antidoom",
+  "wandb_run_name": "kl-sft-r32-e1-openerdiv",
+  "no_save": false
+}

chat_template.jinja

@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if (preserve_thinking is defined and preserve_thinking is true) or (loop.index0 > ns.last_query_index) %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | string if args_value is string else args_value | tojson | safe %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}

config.json

@@ -0,0 +1,119 @@
+{
+    "architectures": [
+        "Qwen3_5MoeForConditionalGeneration"
+    ],
+    "image_token_id": 248056,
+    "model_type": "qwen3_5_moe",
+    "text_config": {
+        "attention_bias": false,
+        "attention_dropout": 0.0,
+        "attn_output_gate": true,
+        "bos_token_id": 248044,
+        "dtype": "bfloat16",
+        "eos_token_id": 248044,
+        "full_attention_interval": 4,
+        "head_dim": 256,
+        "hidden_act": "silu",
+        "hidden_size": 2048,
+        "initializer_range": 0.02,
+        "layer_types": [
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention",
+            "linear_attention",
+            "linear_attention",
+            "linear_attention",
+            "full_attention"
+        ],
+        "linear_conv_kernel_dim": 4,
+        "linear_key_head_dim": 128,
+        "linear_num_key_heads": 16,
+        "linear_num_value_heads": 32,
+        "linear_value_head_dim": 128,
+        "mamba_ssm_dtype": "float32",
+        "max_position_embeddings": 262144,
+        "model_type": "qwen3_5_moe_text",
+        "moe_intermediate_size": 512,
+        "mtp_num_hidden_layers": 1,
+        "mtp_use_dedicated_embeddings": false,
+        "num_attention_heads": 16,
+        "num_experts": 256,
+        "num_experts_per_tok": 8,
+        "num_hidden_layers": 40,
+        "num_key_value_heads": 2,
+        "output_router_logits": false,
+        "pad_token_id": null,
+        "partial_rotary_factor": 0.25,
+        "rms_norm_eps": 1e-06,
+        "rope_parameters": {
+            "mrope_interleaved": true,
+            "mrope_section": [
+                11,
+                11,
+                10
+            ],
+            "partial_rotary_factor": 0.25,
+            "rope_theta": 10000000,
+            "rope_type": "default"
+        },
+        "router_aux_loss_coef": 0.001,
+        "shared_expert_intermediate_size": 512,
+        "tie_word_embeddings": false,
+        "use_cache": true,
+        "vocab_size": 248320
+    },
+    "tie_word_embeddings": false,
+    "transformers_version": "4.57.1",
+    "video_token_id": 248057,
+    "vision_config": {
+        "deepstack_visual_indexes": [],
+        "depth": 27,
+        "hidden_act": "gelu_pytorch_tanh",
+        "hidden_size": 1152,
+        "in_channels": 3,
+        "initializer_range": 0.02,
+        "intermediate_size": 4304,
+        "model_type": "qwen3_5_moe",
+        "num_heads": 16,
+        "num_position_embeddings": 2304,
+        "out_hidden_size": 2048,
+        "patch_size": 16,
+        "spatial_merge_size": 2,
+        "temporal_patch_size": 2
+    },
+    "vision_end_token_id": 248054,
+    "vision_start_token_id": 248053
+}

configuration.json

@@ -0,0 +1 @@
+{"framework":"Pytorch","task":"visual-question-answering"}

generation_config.json

@@ -0,0 +1,12 @@
+{
+    "bos_token_id": 248044,
+    "do_sample": true,
+    "eos_token_id": [
+        248046,
+        248044
+    ],
+    "pad_token_id": 248044,
+    "temperature": 1.0,
+    "top_k": 20,
+    "top_p": 0.95
+}

merge_tools/merge_lora_tensorwise.py

@@ -0,0 +1,425 @@
+#!/usr/bin/env python3
+"""Merge a PEFT LoRA into sharded safetensors with bounded host RAM.
+
+The model is never instantiated.  Each base shard is copied byte-for-byte to a
+temporary output file, then each LoRA target tensor is patched in-place, one
+tensor at a time and in small row tiles.  Completed shards are atomically
+renamed, so ``--resume`` can safely continue after interruption.
+"""
+
+from __future__ import annotations
+
+import argparse
+import gc
+import hashlib
+import json
+import math
+import mmap
+import os
+import platform
+import re
+import shutil
+import struct
+import sys
+import time
+from collections import Counter
+from datetime import datetime, timezone
+from pathlib import Path
+
+import safetensors
+import torch
+from safetensors import safe_open
+
+
+DTYPES = {
+    "BF16": (torch.bfloat16, 2),
+    "F16": (torch.float16, 2),
+    "F32": (torch.float32, 4),
+}
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--base-dir", type=Path, required=True)
+    parser.add_argument("--base-repo", required=True)
+    parser.add_argument("--base-revision", required=True)
+    parser.add_argument("--adapter-dir", type=Path, required=True)
+    parser.add_argument("--output-dir", type=Path, required=True)
+    parser.add_argument("--tile-rows", type=int, default=32)
+    parser.add_argument("--threads", type=int, default=2)
+    parser.add_argument("--expected-targets", type=int, default=190)
+    parser.add_argument("--resume", action="store_true")
+    return parser.parse_args()
+
+
+def sha256_file(path: Path, block_size: int = 8 << 20) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb", buffering=0) as handle:
+        while True:
+            block = handle.read(block_size)
+            if not block:
+                break
+            digest.update(block)
+        try:
+            os.posix_fadvise(handle.fileno(), 0, 0, os.POSIX_FADV_DONTNEED)
+        except (AttributeError, OSError):
+            pass
+    return digest.hexdigest()
+
+
+def atomic_json(path: Path, value: object) -> None:
+    temporary = path.with_name(path.name + ".tmp")
+    temporary.write_text(json.dumps(value, indent=2, sort_keys=True) + "\n")
+    os.replace(temporary, path)
+
+
+def read_safetensors_header(path: Path) -> tuple[int, dict[str, object]]:
+    with path.open("rb") as handle:
+        raw = handle.read(8)
+        if len(raw) != 8:
+            raise ValueError(f"invalid safetensors header in {path}")
+        header_len = struct.unpack("<Q", raw)[0]
+        header = json.loads(handle.read(header_len))
+    return 8 + header_len, header
+
+
+def tensor_entries(header: dict[str, object]) -> dict[str, dict[str, object]]:
+    return {key: value for key, value in header.items() if key != "__metadata__"}
+
+
+def checkpoint_key_for(module_path: str) -> str:
+    match = re.search(r"layers\.\d+\..+$", module_path)
+    if match is None:
+        raise ValueError(f"cannot map adapter module {module_path!r}")
+    return f"model.language_model.{match.group(0)}.weight"
+
+
+def adapter_pairs(adapter_path: Path) -> tuple[dict[str, dict[str, object]], dict[str, object]]:
+    config = json.loads((adapter_path.parent / "adapter_config.json").read_text())
+    rank = int(config["r"])
+    alpha = float(config["lora_alpha"])
+    scaling = alpha / math.sqrt(rank) if config.get("use_rslora") else alpha / rank
+    pairs: dict[str, dict[str, object]] = {}
+    with safe_open(adapter_path, framework="pt", device="cpu") as handle:
+        keys = set(handle.keys())
+        for a_key in sorted(key for key in keys if key.endswith(".lora_A.weight")):
+            module = a_key[: -len(".lora_A.weight")]
+            b_key = module + ".lora_B.weight"
+            if b_key not in keys:
+                raise ValueError(f"missing pair tensor {b_key}")
+            target = checkpoint_key_for(module)
+            if target in pairs:
+                raise ValueError(f"duplicate target {target}")
+            a_slice = handle.get_slice(a_key)
+            b_slice = handle.get_slice(b_key)
+            pairs[target] = {
+                "module": module,
+                "a_key": a_key,
+                "b_key": b_key,
+                "a_shape": list(a_slice.get_shape()),
+                "b_shape": list(b_slice.get_shape()),
+                "a_dtype": str(a_slice.get_dtype()),
+                "b_dtype": str(b_slice.get_dtype()),
+                "scaling": scaling,
+            }
+    return pairs, config
+
+
+def validate_preflight(
+    base_dir: Path,
+    adapter_path: Path,
+    expected_targets: int,
+) -> tuple[
+    dict[str, dict[str, object]],
+    dict[str, object],
+    dict[str, str],
+    dict[str, tuple[int, dict[str, object]]],
+]:
+    index_path = base_dir / "model.safetensors.index.json"
+    index = json.loads(index_path.read_text())
+    weight_map: dict[str, str] = index["weight_map"]
+    pairs, adapter_config = adapter_pairs(adapter_path)
+    if len(pairs) != expected_targets:
+        raise ValueError(f"expected {expected_targets} LoRA targets, found {len(pairs)}")
+
+    headers: dict[str, tuple[int, dict[str, object]]] = {}
+    shard_names = sorted(set(weight_map.values()))
+    for shard_name in shard_names:
+        headers[shard_name] = read_safetensors_header(base_dir / shard_name)
+
+    for target, pair in pairs.items():
+        if target not in weight_map:
+            raise ValueError(f"adapter target missing from base index: {target}")
+        shard_name = weight_map[target]
+        entry = tensor_entries(headers[shard_name][1])[target]
+        base_shape = list(entry["shape"])
+        a_shape = pair["a_shape"]
+        b_shape = pair["b_shape"]
+        if entry["dtype"] != "BF16":
+            raise ValueError(f"target {target} is {entry['dtype']}, expected BF16")
+        if len(base_shape) != 2 or len(a_shape) != 2 or len(b_shape) != 2:
+            raise ValueError(f"target {target} is not a 2-D matrix")
+        if b_shape[0] != base_shape[0] or a_shape[1] != base_shape[1] or b_shape[1] != a_shape[0]:
+            raise ValueError(
+                f"shape mismatch for {target}: base={base_shape}, A={a_shape}, B={b_shape}"
+            )
+    return pairs, adapter_config, weight_map, headers
+
+
+def patch_tensor(
+    mm: mmap.mmap,
+    data_start: int,
+    entry: dict[str, object],
+    a: torch.Tensor,
+    b: torch.Tensor,
+    scaling: float,
+    tile_rows: int,
+) -> dict[str, object]:
+    dtype_name = str(entry["dtype"])
+    if dtype_name not in DTYPES:
+        raise ValueError(f"unsupported target dtype {dtype_name}")
+    dtype, item_size = DTYPES[dtype_name]
+    rows, cols = (int(value) for value in entry["shape"])
+    offset_start, offset_end = (int(value) for value in entry["data_offsets"])
+    if offset_end - offset_start != rows * cols * item_size:
+        raise ValueError("safetensors byte span does not match target shape")
+
+    raw = torch.frombuffer(
+        mm,
+        dtype=dtype,
+        count=rows * cols,
+        offset=data_start + offset_start,
+    ).view(rows, cols)
+    scaled_a = a.float().mul(float(scaling))
+    changed = 0
+    sum_sq = 0.0
+    max_abs = 0.0
+    nonzero_formula = 0
+    for start in range(0, rows, tile_rows):
+        end = min(rows, start + tile_rows)
+        original = raw[start:end].float()
+        formula_delta = b[start:end].float().matmul(scaled_a)
+        merged = (original + formula_delta).to(dtype)
+        changed += int(torch.count_nonzero(merged != raw[start:end]).item())
+        nonzero_formula += int(torch.count_nonzero(formula_delta).item())
+        actual_delta = merged.float().sub(original)
+        if actual_delta.numel():
+            max_abs = max(max_abs, float(actual_delta.abs().max().item()))
+            sum_sq += float(actual_delta.double().square().sum().item())
+        raw[start:end].copy_(merged)
+        del original, formula_delta, merged, actual_delta
+    elements = rows * cols
+    del raw, scaled_a
+    gc.collect()
+    return {
+        "elements": elements,
+        "changed_elements": changed,
+        "changed_fraction": changed / elements,
+        "formula_nonzero_elements": nonzero_formula,
+        "max_abs_effective_bf16_delta": max_abs,
+        "rms_effective_bf16_delta": math.sqrt(sum_sq / elements),
+    }
+
+
+def main() -> None:
+    args = parse_args()
+    if args.tile_rows <= 0 or args.threads <= 0:
+        raise SystemExit("--tile-rows and --threads must be positive")
+    torch.set_num_threads(args.threads)
+    torch.set_num_interop_threads(1)
+
+    base_dir = args.base_dir.resolve()
+    adapter_dir = args.adapter_dir.resolve()
+    output_dir = args.output_dir.resolve()
+    adapter_path = adapter_dir / "adapter_model.safetensors"
+    for required in (adapter_path, adapter_dir / "adapter_config.json", base_dir / "model.safetensors.index.json"):
+        if not required.is_file():
+            raise SystemExit(f"missing required file: {required}")
+
+    output_dir.mkdir(parents=True, exist_ok=True)
+    progress_path = output_dir / ".merge_progress.json"
+    existing = [path for path in output_dir.iterdir() if path.name != ".merge_progress.json"]
+    if existing and not args.resume:
+        raise SystemExit(f"output directory is nonempty; use --resume or a fresh path: {output_dir}")
+    progress = json.loads(progress_path.read_text()) if progress_path.exists() else {"shards": {}}
+
+    pairs, adapter_config, weight_map, headers = validate_preflight(
+        base_dir, adapter_path, args.expected_targets
+    )
+    shard_names = sorted(set(weight_map.values()))
+    targets_by_shard: dict[str, list[str]] = {name: [] for name in shard_names}
+    for target in pairs:
+        targets_by_shard[weight_map[target]].append(target)
+    for shard_name in shard_names:
+        entries = tensor_entries(headers[shard_name][1])
+        targets_by_shard[shard_name].sort(key=lambda key: int(entries[key]["data_offsets"][0]))
+
+    run_identity = {
+        "base_dir": str(base_dir),
+        "base_repo": args.base_repo,
+        "base_revision": args.base_revision,
+        "base_index_sha256": sha256_file(base_dir / "model.safetensors.index.json"),
+        "adapter_dir": str(adapter_dir),
+        "adapter_sha256": sha256_file(adapter_path),
+        "adapter_config_sha256": sha256_file(adapter_dir / "adapter_config.json"),
+        "targets": len(pairs),
+        "tile_rows": args.tile_rows,
+    }
+    if progress.get("run_identity") not in (None, run_identity):
+        raise SystemExit("resume identity differs from the existing merge progress")
+    progress["run_identity"] = run_identity
+    progress.setdefault("started_at", datetime.now(timezone.utc).isoformat())
+    progress.setdefault("shards", {})
+    atomic_json(progress_path, progress)
+
+    target_stats: dict[str, dict[str, object]] = {}
+    started = time.time()
+    with safe_open(adapter_path, framework="pt", device="cpu") as adapter:
+        for shard_number, shard_name in enumerate(shard_names, start=1):
+            source = base_dir / shard_name
+            destination = output_dir / shard_name
+            recorded = progress["shards"].get(shard_name)
+            if args.resume and recorded and destination.is_file():
+                if destination.stat().st_size == recorded["size"] and sha256_file(destination) == recorded["sha256"]:
+                    print(f"[{shard_number:02d}/{len(shard_names)}] verified completed {shard_name}", flush=True)
+                    for target, stats in recorded.get("targets", {}).items():
+                        target_stats[target] = stats
+                    continue
+
+            temporary = output_dir / (shard_name + ".partial")
+            temporary.unlink(missing_ok=True)
+            print(
+                f"[{shard_number:02d}/{len(shard_names)}] copying {shard_name} "
+                f"({source.stat().st_size / (1 << 30):.2f} GiB; {len(targets_by_shard[shard_name])} targets)",
+                flush=True,
+            )
+            shutil.copyfile(source, temporary)
+            shutil.copystat(source, temporary)
+            data_start, header = read_safetensors_header(temporary)
+            entries = tensor_entries(header)
+            shard_target_stats: dict[str, dict[str, object]] = {}
+
+            with temporary.open("r+b", buffering=0) as handle:
+                mm = mmap.mmap(handle.fileno(), 0, access=mmap.ACCESS_WRITE)
+                try:
+                    for target_number, target in enumerate(targets_by_shard[shard_name], start=1):
+                        pair = pairs[target]
+                        a = adapter.get_tensor(pair["a_key"])
+                        b = adapter.get_tensor(pair["b_key"])
+                        stats = patch_tensor(
+                            mm,
+                            data_start,
+                            entries[target],
+                            a,
+                            b,
+                            float(pair["scaling"]),
+                            args.tile_rows,
+                        )
+                        if stats["changed_elements"] == 0:
+                            raise RuntimeError(f"merge rounded to no change for {target}")
+                        stats.update(
+                            {
+                                "module": pair["module"],
+                                "base_shape": list(entries[target]["shape"]),
+                                "a_shape": pair["a_shape"],
+                                "b_shape": pair["b_shape"],
+                                "scaling": pair["scaling"],
+                                "shard": shard_name,
+                            }
+                        )
+                        shard_target_stats[target] = stats
+                        target_stats[target] = stats
+                        del a, b
+                        gc.collect()
+                        mm.flush()
+                        print(
+                            f"    target {target_number:02d}/{len(targets_by_shard[shard_name]):02d} "
+                            f"{target} changed={stats['changed_fraction']:.3%}",
+                            flush=True,
+                        )
+                    mm.flush()
+                    os.fsync(handle.fileno())
+                finally:
+                    mm.close()
+                try:
+                    os.posix_fadvise(handle.fileno(), 0, 0, os.POSIX_FADV_DONTNEED)
+                except (AttributeError, OSError):
+                    pass
+            os.replace(temporary, destination)
+            shard_hash = sha256_file(destination)
+            progress["shards"][shard_name] = {
+                "sha256": shard_hash,
+                "size": destination.stat().st_size,
+                "targets": shard_target_stats,
+                "completed_at": datetime.now(timezone.utc).isoformat(),
+            }
+            atomic_json(progress_path, progress)
+            print(f"    committed sha256={shard_hash[:16]}...", flush=True)
+
+    if len(target_stats) != len(pairs):
+        missing = sorted(set(pairs) - set(target_stats))
+        raise RuntimeError(f"only merged {len(target_stats)}/{len(pairs)} targets: {missing[:3]}")
+
+    for source in base_dir.iterdir():
+        if source.is_file() and not source.name.endswith(".safetensors"):
+            shutil.copy2(source, output_dir / source.name)
+
+    module_counts = Counter(re.search(r"\.([^.]+)$", str(value["module"])).group(1) for value in pairs.values())
+    layer_counts = Counter(
+        int(re.search(r"layers\.(\d+)\.", str(value["module"])).group(1)) for value in pairs.values()
+    )
+    manifest = {
+        "schema_version": 1,
+        "created_at": datetime.now(timezone.utc).isoformat(),
+        "merge_seconds": round(time.time() - started, 3),
+        "base": {
+            "repo": args.base_repo,
+            "revision": args.base_revision,
+            "local_snapshot": str(base_dir),
+            "index_sha256": run_identity["base_index_sha256"],
+            "shards": len(shard_names),
+        },
+        "adapter": {
+            "local_dir": str(adapter_dir),
+            "model_sha256": run_identity["adapter_sha256"],
+            "config_sha256": run_identity["adapter_config_sha256"],
+            "training_args_sha256": (
+                sha256_file(adapter_dir / "training_args.json")
+                if (adapter_dir / "training_args.json").is_file()
+                else None
+            ),
+            "config": adapter_config,
+        },
+        "merge": {
+            "formula": "BF16(W_FP32 + B_FP32 @ (A_FP32 * (lora_alpha / r)))",
+            "strategy": "one target tensor at a time, row-tiled, in-place in a copied shard",
+            "tile_rows": args.tile_rows,
+            "threads": args.threads,
+            "target_count": len(pairs),
+            "module_counts": dict(sorted(module_counts.items())),
+            "layer_counts": {str(key): layer_counts[key] for key in sorted(layer_counts)},
+        },
+        "target_stats": target_stats,
+        "shards": {
+            name: {"sha256": value["sha256"], "size": value["size"]}
+            for name, value in progress["shards"].items()
+        },
+        "environment": {
+            "python": sys.version,
+            "platform": platform.platform(),
+            "torch": torch.__version__,
+            "safetensors": safetensors.__version__,
+            "script_sha256": sha256_file(Path(__file__).resolve()),
+        },
+    }
+    atomic_json(output_dir / "merge_manifest.json", manifest)
+    (output_dir / "MERGE_COMPLETE").write_text(datetime.now(timezone.utc).isoformat() + "\n")
+    progress_path.unlink(missing_ok=True)
+    print(
+        f"MERGE COMPLETE: {len(target_stats)} targets, {len(shard_names)} shards -> {output_dir}",
+        flush=True,
+    )
+
+
+if __name__ == "__main__":
+    main()

merge_tools/verify_lora_tensorwise.py

@@ -0,0 +1,359 @@
+#!/usr/bin/env python3
+"""Certify a tensor-wise LoRA merge without loading the model.
+
+For every LoRA target, recompute the FP32 LoRA formula in small row tiles and
+require bit-exact BF16 equality with the merged tensor.  For every byte outside
+the 190 target tensor ranges, require exact equality with the pinned base
+checkpoint.  Optionally require every target to differ from an older merge.
+"""
+
+from __future__ import annotations
+
+import argparse
+import gc
+import hashlib
+import json
+import math
+import mmap
+import os
+import re
+import struct
+import time
+from collections import Counter, defaultdict
+from datetime import datetime, timezone
+from pathlib import Path
+
+import torch
+from safetensors import safe_open
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--base-dir", type=Path, required=True)
+    parser.add_argument("--adapter-dir", type=Path, required=True)
+    parser.add_argument("--merged-dir", type=Path, required=True)
+    parser.add_argument("--old-merged-dir", type=Path)
+    parser.add_argument("--tile-rows", type=int, default=32)
+    parser.add_argument("--threads", type=int, default=2)
+    parser.add_argument("--compare-block-mib", type=int, default=8)
+    parser.add_argument("--expected-targets", type=int, default=190)
+    return parser.parse_args()
+
+
+def sha256_file(path: Path, block_size: int = 8 << 20) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb", buffering=0) as handle:
+        while True:
+            block = handle.read(block_size)
+            if not block:
+                break
+            digest.update(block)
+        try:
+            os.posix_fadvise(handle.fileno(), 0, 0, os.POSIX_FADV_DONTNEED)
+        except (AttributeError, OSError):
+            pass
+    return digest.hexdigest()
+
+
+def atomic_json(path: Path, value: object) -> None:
+    temporary = path.with_name(path.name + ".tmp")
+    temporary.write_text(json.dumps(value, indent=2, sort_keys=True) + "\n")
+    os.replace(temporary, path)
+
+
+def read_header(path: Path) -> tuple[int, bytes, dict[str, object]]:
+    with path.open("rb") as handle:
+        prefix = handle.read(8)
+        if len(prefix) != 8:
+            raise ValueError(f"invalid safetensors file: {path}")
+        header_len = struct.unpack("<Q", prefix)[0]
+        header_bytes = handle.read(header_len)
+    return 8 + header_len, prefix + header_bytes, json.loads(header_bytes)
+
+
+def entries(header: dict[str, object]) -> dict[str, dict[str, object]]:
+    return {key: value for key, value in header.items() if key != "__metadata__"}
+
+
+def checkpoint_key_for(module_path: str) -> str:
+    match = re.search(r"layers\.\d+\..+$", module_path)
+    if match is None:
+        raise ValueError(f"cannot map adapter module {module_path}")
+    return f"model.language_model.{match.group(0)}.weight"
+
+
+def discover_pairs(adapter_dir: Path) -> tuple[dict[str, dict[str, object]], float]:
+    config = json.loads((adapter_dir / "adapter_config.json").read_text())
+    rank = int(config["r"])
+    alpha = float(config["lora_alpha"])
+    scaling = alpha / math.sqrt(rank) if config.get("use_rslora") else alpha / rank
+    pairs: dict[str, dict[str, object]] = {}
+    with safe_open(adapter_dir / "adapter_model.safetensors", framework="pt", device="cpu") as adapter:
+        keys = set(adapter.keys())
+        for a_key in sorted(key for key in keys if key.endswith(".lora_A.weight")):
+            module = a_key[: -len(".lora_A.weight")]
+            b_key = module + ".lora_B.weight"
+            if b_key not in keys:
+                raise ValueError(f"missing {b_key}")
+            target = checkpoint_key_for(module)
+            pairs[target] = {"module": module, "a_key": a_key, "b_key": b_key}
+    return pairs, scaling
+
+
+def compare_range(
+    base_handle,
+    merged_handle,
+    offset: int,
+    length: int,
+    block_size: int,
+) -> None:
+    base_handle.seek(offset)
+    merged_handle.seek(offset)
+    remaining = length
+    while remaining:
+        wanted = min(block_size, remaining)
+        base_block = base_handle.read(wanted)
+        merged_block = merged_handle.read(wanted)
+        if len(base_block) != wanted or len(merged_block) != wanted:
+            raise RuntimeError(f"short read while comparing at byte {offset + length - remaining}")
+        if base_block != merged_block:
+            start = offset + length - remaining
+            first = next(i for i, (left, right) in enumerate(zip(base_block, merged_block)) if left != right)
+            raise RuntimeError(f"non-target byte changed at absolute offset {start + first}")
+        remaining -= wanted
+
+
+def read_bf16_matrix(mm: mmap.mmap, data_start: int, entry: dict[str, object]) -> torch.Tensor:
+    if entry["dtype"] != "BF16":
+        raise ValueError(f"expected BF16, got {entry['dtype']}")
+    rows, columns = (int(value) for value in entry["shape"])
+    start, end = (int(value) for value in entry["data_offsets"])
+    if end - start != rows * columns * 2:
+        raise ValueError("invalid BF16 tensor span")
+    return torch.frombuffer(
+        mm,
+        dtype=torch.bfloat16,
+        count=rows * columns,
+        offset=data_start + start,
+    ).view(rows, columns)
+
+
+def main() -> None:
+    args = parse_args()
+    if min(args.tile_rows, args.threads, args.compare_block_mib) <= 0:
+        raise SystemExit("tile rows, threads, and compare block must be positive")
+    torch.set_num_threads(args.threads)
+    torch.set_num_interop_threads(1)
+    block_size = args.compare_block_mib << 20
+
+    base_dir = args.base_dir.resolve()
+    adapter_dir = args.adapter_dir.resolve()
+    merged_dir = args.merged_dir.resolve()
+    old_dir = args.old_merged_dir.resolve() if args.old_merged_dir else None
+    manifest_path = merged_dir / "merge_manifest.json"
+    if not manifest_path.is_file() or not (merged_dir / "MERGE_COMPLETE").is_file():
+        raise SystemExit("merged artifact lacks merge_manifest.json or MERGE_COMPLETE")
+    if list(merged_dir.glob("*.partial")):
+        raise SystemExit("merged artifact still contains partial shard files")
+
+    manifest = json.loads(manifest_path.read_text())
+    adapter_hash = sha256_file(adapter_dir / "adapter_model.safetensors")
+    if adapter_hash != manifest["adapter"]["model_sha256"]:
+        raise SystemExit("adapter hash differs from merge manifest")
+    if sha256_file(base_dir / "model.safetensors.index.json") != manifest["base"]["index_sha256"]:
+        raise SystemExit("base index hash differs from merge manifest")
+
+    base_index = json.loads((base_dir / "model.safetensors.index.json").read_text())
+    merged_index = json.loads((merged_dir / "model.safetensors.index.json").read_text())
+    if merged_index != base_index:
+        raise SystemExit("merged model index is not identical to the pinned base index")
+    weight_map: dict[str, str] = base_index["weight_map"]
+    shard_names = sorted(set(weight_map.values()))
+    if len(shard_names) != 26:
+        raise SystemExit(f"expected 26 base shards, found {len(shard_names)}")
+    pairs, scaling = discover_pairs(adapter_dir)
+    if len(pairs) != args.expected_targets:
+        raise SystemExit(f"expected {args.expected_targets} targets, found {len(pairs)}")
+    if set(pairs) - set(weight_map):
+        raise SystemExit(f"adapter targets missing from base: {sorted(set(pairs)-set(weight_map))[:3]}")
+
+    targets_by_shard: dict[str, list[str]] = defaultdict(list)
+    for target in pairs:
+        targets_by_shard[weight_map[target]].append(target)
+
+    target_reports: dict[str, dict[str, object]] = {}
+    unchanged_bytes = 0
+    started = time.time()
+    with safe_open(adapter_dir / "adapter_model.safetensors", framework="pt", device="cpu") as adapter:
+        for shard_number, shard_name in enumerate(shard_names, start=1):
+            base_path = base_dir / shard_name
+            merged_path = merged_dir / shard_name
+            if not merged_path.is_file() or merged_path.stat().st_size != base_path.stat().st_size:
+                raise RuntimeError(f"missing or wrong-sized merged shard {shard_name}")
+            expected_hash = manifest["shards"][shard_name]["sha256"]
+            actual_hash = sha256_file(merged_path)
+            if actual_hash != expected_hash:
+                raise RuntimeError(f"merged shard hash mismatch: {shard_name}")
+
+            base_data_start, base_header_bytes, base_header = read_header(base_path)
+            merged_data_start, merged_header_bytes, merged_header = read_header(merged_path)
+            if base_data_start != merged_data_start or base_header_bytes != merged_header_bytes:
+                raise RuntimeError(f"safetensors header changed in {shard_name}")
+            base_entries = entries(base_header)
+            merged_entries = entries(merged_header)
+            if base_entries != merged_entries:
+                raise RuntimeError(f"tensor index changed in {shard_name}")
+
+            target_ranges = []
+            for target in targets_by_shard[shard_name]:
+                start, end = (int(value) for value in base_entries[target]["data_offsets"])
+                target_ranges.append((base_data_start + start, base_data_start + end, target))
+            target_ranges.sort()
+            cursor = 0
+            with base_path.open("rb", buffering=0) as base_handle, merged_path.open("rb", buffering=0) as merged_handle:
+                for start, end, _ in target_ranges:
+                    compare_range(base_handle, merged_handle, cursor, start - cursor, block_size)
+                    unchanged_bytes += start - cursor
+                    cursor = end
+                compare_range(base_handle, merged_handle, cursor, base_path.stat().st_size - cursor, block_size)
+                unchanged_bytes += base_path.stat().st_size - cursor
+                for handle in (base_handle, merged_handle):
+                    try:
+                        os.posix_fadvise(handle.fileno(), 0, 0, os.POSIX_FADV_DONTNEED)
+                    except (AttributeError, OSError):
+                        pass
+
+            old_path = old_dir / shard_name if old_dir else None
+            with base_path.open("rb", buffering=0) as base_file, merged_path.open("rb", buffering=0) as merged_file:
+                base_mm = mmap.mmap(base_file.fileno(), 0, access=mmap.ACCESS_READ)
+                merged_mm = mmap.mmap(merged_file.fileno(), 0, access=mmap.ACCESS_READ)
+                old_file = old_path.open("rb", buffering=0) if old_path and old_path.is_file() else None
+                old_mm = mmap.mmap(old_file.fileno(), 0, access=mmap.ACCESS_READ) if old_file else None
+                old_entries = entries(read_header(old_path)[2]) if old_path and old_path.is_file() else None
+                try:
+                    for target_number, target in enumerate(targets_by_shard[shard_name], start=1):
+                        pair = pairs[target]
+                        a = adapter.get_tensor(pair["a_key"])
+                        b = adapter.get_tensor(pair["b_key"])
+                        base_tensor = read_bf16_matrix(base_mm, base_data_start, base_entries[target])
+                        merged_tensor = read_bf16_matrix(merged_mm, merged_data_start, merged_entries[target])
+                        old_tensor = (
+                            read_bf16_matrix(old_mm, read_header(old_path)[0], old_entries[target])
+                            if old_mm is not None and old_entries is not None
+                            else None
+                        )
+                        scaled_a = a.float().mul(scaling)
+                        changed = 0
+                        differs_old = 0
+                        max_abs = 0.0
+                        sum_sq = 0.0
+                        rows, columns = base_tensor.shape
+                        for start in range(0, rows, args.tile_rows):
+                            end = min(rows, start + args.tile_rows)
+                            expected = (
+                                base_tensor[start:end].float()
+                                + b[start:end].float().matmul(scaled_a)
+                            ).to(torch.bfloat16)
+                            if not torch.equal(expected, merged_tensor[start:end]):
+                                mismatches = int(torch.count_nonzero(expected != merged_tensor[start:end]).item())
+                                raise RuntimeError(f"formula mismatch for {target}: {mismatches} elements")
+                            changed += int(torch.count_nonzero(merged_tensor[start:end] != base_tensor[start:end]).item())
+                            if old_tensor is not None:
+                                differs_old += int(torch.count_nonzero(merged_tensor[start:end] != old_tensor[start:end]).item())
+                            actual = merged_tensor[start:end].float() - base_tensor[start:end].float()
+                            max_abs = max(max_abs, float(actual.abs().max().item()))
+                            sum_sq += float(actual.double().square().sum().item())
+                            del expected, actual
+                        if changed == 0:
+                            raise RuntimeError(f"merged target is identical to base: {target}")
+                        if old_tensor is not None and differs_old == 0:
+                            raise RuntimeError(f"round-2 merged target is identical to round 1: {target}")
+                        elements = rows * columns
+                        target_reports[target] = {
+                            "module": pair["module"],
+                            "shard": shard_name,
+                            "shape": [rows, columns],
+                            "changed_from_base_elements": changed,
+                            "changed_from_base_fraction": changed / elements,
+                            "changed_from_round1_elements": differs_old if old_tensor is not None else None,
+                            "max_abs_effective_bf16_delta": max_abs,
+                            "rms_effective_bf16_delta": math.sqrt(sum_sq / elements),
+                            "formula_bit_exact": True,
+                        }
+                        del a, b, base_tensor, merged_tensor, old_tensor, scaled_a
+                        gc.collect()
+                        print(
+                            f"[{shard_number:02d}/{len(shard_names)}] target "
+                            f"{target_number:02d}/{len(targets_by_shard[shard_name]):02d} verified {target}",
+                            flush=True,
+                        )
+                finally:
+                    base_mm.close()
+                    merged_mm.close()
+                    if old_mm is not None:
+                        old_mm.close()
+                    if old_file is not None:
+                        old_file.close()
+            print(
+                f"[{shard_number:02d}/{len(shard_names)}] {shard_name}: hash, non-target bytes, and formulas verified",
+                flush=True,
+            )
+
+    if set(target_reports) != set(pairs):
+        raise RuntimeError(f"verified {len(target_reports)}/{len(pairs)} targets")
+    module_counts = Counter(re.search(r"\.([^.]+)$", value["module"]).group(1) for value in pairs.values())
+    layer_counts = Counter(int(re.search(r"layers\.(\d+)\.", value["module"]).group(1)) for value in pairs.values())
+    expected_module_counts = {
+        "in_proj_a": 30,
+        "in_proj_b": 30,
+        "in_proj_qkv": 30,
+        "in_proj_z": 30,
+        "out_proj": 30,
+        "k_proj": 10,
+        "o_proj": 10,
+        "q_proj": 10,
+        "v_proj": 10,
+    }
+    if dict(module_counts) != expected_module_counts:
+        raise RuntimeError(f"unexpected target-module coverage: {dict(module_counts)}")
+    if set(layer_counts) != set(range(40)):
+        raise RuntimeError(f"not all 40 transformer layers are targeted: {dict(layer_counts)}")
+
+    metadata_loads: dict[str, str] = {}
+    from transformers import AutoConfig, AutoProcessor, AutoTokenizer
+
+    config = AutoConfig.from_pretrained(merged_dir, local_files_only=True)
+    metadata_loads["config_class"] = type(config).__name__
+    tokenizer = AutoTokenizer.from_pretrained(merged_dir, local_files_only=True)
+    metadata_loads["tokenizer_class"] = type(tokenizer).__name__
+    processor = AutoProcessor.from_pretrained(merged_dir, local_files_only=True)
+    metadata_loads["processor_class"] = type(processor).__name__
+
+    report = {
+        "schema_version": 1,
+        "verified_at": datetime.now(timezone.utc).isoformat(),
+        "verification_seconds": round(time.time() - started, 3),
+        "base_revision": manifest["base"]["revision"],
+        "adapter_sha256": adapter_hash,
+        "merged_shards": len(shard_names),
+        "base_tensor_count": len(weight_map),
+        "target_tensor_count": len(target_reports),
+        "non_target_tensor_count": len(weight_map) - len(target_reports),
+        "unchanged_bytes_compared": unchanged_bytes,
+        "all_non_target_bytes_identical": True,
+        "all_targets_formula_bit_exact": True,
+        "all_targets_distinct_from_base": True,
+        "all_targets_distinct_from_round1": old_dir is not None,
+        "module_counts": dict(sorted(module_counts.items())),
+        "layer_counts": {str(key): layer_counts[key] for key in sorted(layer_counts)},
+        "metadata_loads": metadata_loads,
+        "target_reports": target_reports,
+    }
+    atomic_json(merged_dir / "merge_verification.json", report)
+    (merged_dir / "VERIFIED").write_text(datetime.now(timezone.utc).isoformat() + "\n")
+    print(json.dumps({key: value for key, value in report.items() if key != "target_reports"}, indent=2))
+    print(f"VERIFIED: {merged_dir}", flush=True)
+
+
+if __name__ == "__main__":
+    main()

model-00001-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f29ad3f6b31540418066e66ea9a017c6f112d5ab185e57ba2f41c9fb705978e
+size 3996199712

model-00002-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02edba7a7845c762b9da61d736b8404d680c33c96687f957e767b44f7cff954c
+size 1284907696

model-00003-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39b9acfcfeb1d9062e449c61fe9a0e6975c40e28dae4f9b476358f6e5ccfeb74
+size 3357898360

model-00004-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:972bae2f88c1369355ef84a863ddfa8a3c66b2029d54fdd4f3299f6b87475184
+size 3370808712

model-00005-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00609f3287a54c589843e5e41d6f1d83f2f4f83af49fab6192c3c207e74b7062
+size 3357898360

model-00006-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6175f8ffe687723d9b140fb9ca2eeb6c5705d46d0fd8b6afe291703db3f3e930
+size 3959424904

model-00007-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab54c82bf2cb55dda4994c6651b95e6a41a1ac834f6ee883812743a3dc21fe99
+size 1096788232

model-00008-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a3b4b8d0c96f9b42ed0ea1ca4d460a06a6032c7286a2a090060aac7605f8023
+size 3946842008

model-00009-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d940216c8c6c11cad0b32390a372c6b73fe420b86f04ae2bf4be18b7ce7818ec
+size 1096460848

model-00010-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:001f7913d74dab95f9687006c65efbb958d652be7f0ecf3f7bda757cbd90f1f9
+size 3946841992

model-00011-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a089da7bc158b113ed15a7f7d515f9311da156532eeab17ff6d7e4cef9679b4
+size 1096460752

model-00012-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d51714d98723f308ec5bb3137514951fa202abfc5e01b66c75181dd39804688a
+size 3409971080

model-00013-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36a6e8f572fca3310255fb001f6b116ae90781e2f4798eee6dc5a79ee92fcf8f
+size 1633331664

model-00014-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c03d665943d7dbbcc5016e9674180fb642f67c2cec4c768dff85e2f00ab0cc68
+size 3422553872

model-00015-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a54d60d24c7234952847c71c9c8eaa780136cb691d0e1254aa93b3fe11f922d1
+size 1633659224

model-00016-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:856939b826bcae35bd6a88a0afbc04d838306712b3731b947927e59afee562d9
+size 3946842136

model-00017-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35058ac91b5652a294842a67d970e738c8ccb4b409e875565eb56199d3aa997a
+size 1096460608

model-00018-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f02e3886d3ec19bf5e8ee4401f0b038ee4908ec277c9562262a019a55bc1984
+size 3946841992

model-00019-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e284dc89e707c65639714385c13b0a0148db71b685ae7ace9dcace0bc07dd5b6
+size 1096460808

model-00020-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:365eb7ffa4efb30a9605eec3212462df30c1f92f0bc31b23ad88447601ff3026
+size 3409971072

model-00021-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d74daf37694a5d997cafe1d05fddc64aad7a868a210675c78e5e9d3f87150ab9
+size 1633331744

model-00022-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9059e5816bb428e302a5fc575ad55fd3ea1c77ececde84f599d496834f8218b
+size 3370808752

model-00023-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4261ba26ddd9b875d04bed2e35e5d93ba4cdda923d4b51f1bd476c310ad0acb
+size 3357898392

model-00024-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd381409ec88968f177fc2968c1b311ec1bad48a8efaace6104918c439304758
+size 3370808752

model-00025-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4dc3961a2e10e6b01da2f19bf546b60e5fe8a520d219231b8ab38b618865fb13
+size 3832888256

model-00026-of-00026.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a30d2c6c2a09f92bce458c4d42fbf4bc3410062d25f1a2d020ac42035c107b1a
+size 2231416848

model.safetensors.index.json

@@ -0,0 +1,1052 @@
+{
+  "metadata": {
+    "total_size": 71903645408.0
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.embed_tokens.weight": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.0.input_layernorm.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.linear_attn.A_log": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.linear_attn.conv1d.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.linear_attn.dt_bias": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.linear_attn.in_proj_a.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.linear_attn.in_proj_b.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.linear_attn.in_proj_qkv.weight": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.0.linear_attn.in_proj_z.weight": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.0.linear_attn.norm.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.linear_attn.out_proj.weight": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.0.mlp.experts.down_proj": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.mlp.experts.gate_up_proj": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.0.mlp.gate.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.mlp.shared_expert.down_proj.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.mlp.shared_expert.gate_proj.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.mlp.shared_expert.up_proj.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.mlp.shared_expert_gate.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.0.post_attention_layernorm.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.input_layernorm.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.linear_attn.A_log": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.linear_attn.conv1d.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.linear_attn.dt_bias": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.linear_attn.in_proj_a.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.linear_attn.in_proj_b.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.linear_attn.in_proj_qkv.weight": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.1.linear_attn.in_proj_z.weight": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.1.linear_attn.norm.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.linear_attn.out_proj.weight": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.1.mlp.experts.down_proj": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.1.mlp.experts.gate_up_proj": "model-00001-of-00026.safetensors",
+    "model.language_model.layers.1.mlp.gate.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.mlp.shared_expert.down_proj.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.mlp.shared_expert.gate_proj.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.mlp.shared_expert.up_proj.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.mlp.shared_expert_gate.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.1.post_attention_layernorm.weight": "model-00002-of-00026.safetensors",
+    "model.language_model.layers.10.input_layernorm.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.linear_attn.A_log": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.linear_attn.conv1d.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.linear_attn.dt_bias": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.linear_attn.in_proj_a.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.linear_attn.in_proj_b.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.linear_attn.in_proj_qkv.weight": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.10.linear_attn.in_proj_z.weight": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.10.linear_attn.norm.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.linear_attn.out_proj.weight": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.10.mlp.experts.down_proj": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.10.mlp.experts.gate_up_proj": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.mlp.gate.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.mlp.shared_expert.down_proj.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.mlp.shared_expert.gate_proj.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.mlp.shared_expert.up_proj.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.mlp.shared_expert_gate.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.10.post_attention_layernorm.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.11.input_layernorm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.mlp.experts.down_proj": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.mlp.experts.gate_up_proj": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.11.mlp.gate.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.mlp.shared_expert.down_proj.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.mlp.shared_expert.gate_proj.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.mlp.shared_expert.up_proj.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.mlp.shared_expert_gate.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.post_attention_layernorm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.self_attn.k_norm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.self_attn.k_proj.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.11.self_attn.o_proj.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.11.self_attn.q_norm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.11.self_attn.q_proj.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.11.self_attn.v_proj.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.12.input_layernorm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.linear_attn.A_log": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.linear_attn.conv1d.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.linear_attn.dt_bias": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.linear_attn.in_proj_a.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.linear_attn.in_proj_b.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.linear_attn.in_proj_qkv.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.12.linear_attn.in_proj_z.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.12.linear_attn.norm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.linear_attn.out_proj.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.12.mlp.experts.down_proj": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.mlp.experts.gate_up_proj": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.12.mlp.gate.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.mlp.shared_expert.down_proj.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.mlp.shared_expert.gate_proj.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.mlp.shared_expert.up_proj.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.mlp.shared_expert_gate.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.12.post_attention_layernorm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.input_layernorm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.linear_attn.A_log": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.linear_attn.conv1d.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.linear_attn.dt_bias": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.linear_attn.in_proj_a.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.linear_attn.in_proj_b.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.linear_attn.in_proj_qkv.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.13.linear_attn.in_proj_z.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.13.linear_attn.norm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.linear_attn.out_proj.weight": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.13.mlp.experts.down_proj": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.13.mlp.experts.gate_up_proj": "model-00008-of-00026.safetensors",
+    "model.language_model.layers.13.mlp.gate.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.mlp.shared_expert.down_proj.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.mlp.shared_expert.gate_proj.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.mlp.shared_expert.up_proj.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.mlp.shared_expert_gate.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.13.post_attention_layernorm.weight": "model-00009-of-00026.safetensors",
+    "model.language_model.layers.14.input_layernorm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.linear_attn.A_log": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.linear_attn.conv1d.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.linear_attn.dt_bias": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.linear_attn.in_proj_a.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.linear_attn.in_proj_b.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.linear_attn.in_proj_qkv.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.14.linear_attn.in_proj_z.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.14.linear_attn.norm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.linear_attn.out_proj.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.14.mlp.experts.down_proj": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.14.mlp.experts.gate_up_proj": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.14.mlp.gate.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.mlp.shared_expert.down_proj.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.mlp.shared_expert.gate_proj.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.mlp.shared_expert.up_proj.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.mlp.shared_expert_gate.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.14.post_attention_layernorm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.input_layernorm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.mlp.experts.down_proj": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.mlp.experts.gate_up_proj": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.15.mlp.gate.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.mlp.shared_expert.down_proj.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.mlp.shared_expert.gate_proj.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.mlp.shared_expert.up_proj.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.mlp.shared_expert_gate.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.post_attention_layernorm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.self_attn.k_norm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.self_attn.k_proj.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.15.self_attn.o_proj.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.15.self_attn.q_norm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.15.self_attn.q_proj.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.15.self_attn.v_proj.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.16.input_layernorm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.linear_attn.A_log": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.linear_attn.conv1d.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.linear_attn.dt_bias": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.linear_attn.in_proj_a.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.linear_attn.in_proj_b.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.linear_attn.in_proj_qkv.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.16.linear_attn.in_proj_z.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.16.linear_attn.norm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.linear_attn.out_proj.weight": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.16.mlp.experts.down_proj": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.mlp.experts.gate_up_proj": "model-00010-of-00026.safetensors",
+    "model.language_model.layers.16.mlp.gate.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.mlp.shared_expert.down_proj.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.mlp.shared_expert.gate_proj.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.mlp.shared_expert.up_proj.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.mlp.shared_expert_gate.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.16.post_attention_layernorm.weight": "model-00011-of-00026.safetensors",
+    "model.language_model.layers.17.input_layernorm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.linear_attn.A_log": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.linear_attn.conv1d.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.linear_attn.dt_bias": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.linear_attn.in_proj_a.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.linear_attn.in_proj_b.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.linear_attn.in_proj_qkv.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.17.linear_attn.in_proj_z.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.17.linear_attn.norm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.linear_attn.out_proj.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.17.mlp.experts.down_proj": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.17.mlp.experts.gate_up_proj": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.17.mlp.gate.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.mlp.shared_expert.down_proj.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.mlp.shared_expert.gate_proj.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.mlp.shared_expert.up_proj.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.mlp.shared_expert_gate.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.17.post_attention_layernorm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.input_layernorm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.linear_attn.A_log": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.linear_attn.conv1d.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.linear_attn.dt_bias": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.linear_attn.in_proj_a.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.linear_attn.in_proj_b.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.linear_attn.in_proj_qkv.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.18.linear_attn.in_proj_z.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.18.linear_attn.norm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.linear_attn.out_proj.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.18.mlp.experts.down_proj": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.18.mlp.experts.gate_up_proj": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.mlp.gate.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.mlp.shared_expert.down_proj.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.mlp.shared_expert.gate_proj.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.mlp.shared_expert.up_proj.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.mlp.shared_expert_gate.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.18.post_attention_layernorm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.input_layernorm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.mlp.experts.down_proj": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.mlp.experts.gate_up_proj": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.19.mlp.gate.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.mlp.shared_expert.down_proj.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.mlp.shared_expert.gate_proj.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.mlp.shared_expert.up_proj.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.mlp.shared_expert_gate.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.post_attention_layernorm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.self_attn.k_norm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.self_attn.k_proj.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.19.self_attn.o_proj.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.19.self_attn.q_norm.weight": "model-00013-of-00026.safetensors",
+    "model.language_model.layers.19.self_attn.q_proj.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.19.self_attn.v_proj.weight": "model-00012-of-00026.safetensors",
+    "model.language_model.layers.2.input_layernorm.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.linear_attn.A_log": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.linear_attn.conv1d.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.linear_attn.dt_bias": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.linear_attn.in_proj_a.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.linear_attn.in_proj_b.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.linear_attn.in_proj_qkv.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.linear_attn.in_proj_z.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.linear_attn.norm.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.linear_attn.out_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.mlp.experts.down_proj": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.mlp.experts.gate_up_proj": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.mlp.gate.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.mlp.shared_expert.down_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.mlp.shared_expert.gate_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.mlp.shared_expert.up_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.mlp.shared_expert_gate.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.2.post_attention_layernorm.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.20.input_layernorm.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.linear_attn.A_log": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.linear_attn.conv1d.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.linear_attn.dt_bias": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.linear_attn.in_proj_a.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.linear_attn.in_proj_b.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.linear_attn.in_proj_qkv.weight": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.20.linear_attn.in_proj_z.weight": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.20.linear_attn.norm.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.linear_attn.out_proj.weight": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.20.mlp.experts.down_proj": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.20.mlp.experts.gate_up_proj": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.20.mlp.gate.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.mlp.shared_expert.down_proj.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.mlp.shared_expert.gate_proj.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.mlp.shared_expert.up_proj.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.mlp.shared_expert_gate.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.20.post_attention_layernorm.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.input_layernorm.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.linear_attn.A_log": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.linear_attn.conv1d.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.linear_attn.dt_bias": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.linear_attn.in_proj_a.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.linear_attn.in_proj_b.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.linear_attn.in_proj_qkv.weight": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.21.linear_attn.in_proj_z.weight": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.21.linear_attn.norm.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.linear_attn.out_proj.weight": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.21.mlp.experts.down_proj": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.21.mlp.experts.gate_up_proj": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.mlp.gate.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.mlp.shared_expert.down_proj.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.mlp.shared_expert.gate_proj.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.mlp.shared_expert.up_proj.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.mlp.shared_expert_gate.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.21.post_attention_layernorm.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.input_layernorm.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.linear_attn.A_log": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.linear_attn.conv1d.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.linear_attn.dt_bias": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.linear_attn.in_proj_a.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.linear_attn.in_proj_b.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.linear_attn.in_proj_qkv.weight": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.22.linear_attn.in_proj_z.weight": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.22.linear_attn.norm.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.linear_attn.out_proj.weight": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.22.mlp.experts.down_proj": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.mlp.experts.gate_up_proj": "model-00014-of-00026.safetensors",
+    "model.language_model.layers.22.mlp.gate.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.mlp.shared_expert.down_proj.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.mlp.shared_expert.gate_proj.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.mlp.shared_expert.up_proj.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.mlp.shared_expert_gate.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.22.post_attention_layernorm.weight": "model-00015-of-00026.safetensors",
+    "model.language_model.layers.23.input_layernorm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.23.mlp.experts.down_proj": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.23.mlp.experts.gate_up_proj": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.23.mlp.gate.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.23.mlp.shared_expert.down_proj.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.23.mlp.shared_expert.gate_proj.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.23.mlp.shared_expert.up_proj.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.23.mlp.shared_expert_gate.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.23.post_attention_layernorm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.23.self_attn.k_norm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.23.self_attn.k_proj.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.23.self_attn.o_proj.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.23.self_attn.q_norm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.23.self_attn.q_proj.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.23.self_attn.v_proj.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.24.input_layernorm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.linear_attn.A_log": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.linear_attn.conv1d.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.linear_attn.dt_bias": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.linear_attn.in_proj_a.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.linear_attn.in_proj_b.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.linear_attn.in_proj_qkv.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.24.linear_attn.in_proj_z.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.24.linear_attn.norm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.linear_attn.out_proj.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.24.mlp.experts.down_proj": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.24.mlp.experts.gate_up_proj": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.24.mlp.gate.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.mlp.shared_expert.down_proj.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.mlp.shared_expert.gate_proj.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.mlp.shared_expert.up_proj.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.mlp.shared_expert_gate.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.24.post_attention_layernorm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.input_layernorm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.linear_attn.A_log": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.linear_attn.conv1d.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.linear_attn.dt_bias": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.linear_attn.in_proj_a.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.linear_attn.in_proj_b.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.linear_attn.in_proj_qkv.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.25.linear_attn.in_proj_z.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.25.linear_attn.norm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.linear_attn.out_proj.weight": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.25.mlp.experts.down_proj": "model-00016-of-00026.safetensors",
+    "model.language_model.layers.25.mlp.experts.gate_up_proj": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.mlp.gate.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.mlp.shared_expert.down_proj.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.mlp.shared_expert.gate_proj.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.mlp.shared_expert.up_proj.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.mlp.shared_expert_gate.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.25.post_attention_layernorm.weight": "model-00017-of-00026.safetensors",
+    "model.language_model.layers.26.input_layernorm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.linear_attn.A_log": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.linear_attn.conv1d.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.linear_attn.dt_bias": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.linear_attn.in_proj_a.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.linear_attn.in_proj_b.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.linear_attn.in_proj_qkv.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.26.linear_attn.in_proj_z.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.26.linear_attn.norm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.linear_attn.out_proj.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.26.mlp.experts.down_proj": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.mlp.experts.gate_up_proj": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.26.mlp.gate.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.mlp.shared_expert.down_proj.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.mlp.shared_expert.gate_proj.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.mlp.shared_expert.up_proj.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.mlp.shared_expert_gate.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.26.post_attention_layernorm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.input_layernorm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.mlp.experts.down_proj": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.mlp.experts.gate_up_proj": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.27.mlp.gate.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.mlp.shared_expert.down_proj.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.mlp.shared_expert.gate_proj.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.mlp.shared_expert.up_proj.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.mlp.shared_expert_gate.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.post_attention_layernorm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.self_attn.k_norm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.self_attn.k_proj.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.27.self_attn.o_proj.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.27.self_attn.q_norm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.27.self_attn.q_proj.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.27.self_attn.v_proj.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.28.input_layernorm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.linear_attn.A_log": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.linear_attn.conv1d.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.linear_attn.dt_bias": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.linear_attn.in_proj_a.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.linear_attn.in_proj_b.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.linear_attn.in_proj_qkv.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.28.linear_attn.in_proj_z.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.28.linear_attn.norm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.linear_attn.out_proj.weight": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.28.mlp.experts.down_proj": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.28.mlp.experts.gate_up_proj": "model-00018-of-00026.safetensors",
+    "model.language_model.layers.28.mlp.gate.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.mlp.shared_expert.down_proj.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.mlp.shared_expert.gate_proj.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.mlp.shared_expert.up_proj.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.mlp.shared_expert_gate.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.28.post_attention_layernorm.weight": "model-00019-of-00026.safetensors",
+    "model.language_model.layers.29.input_layernorm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.linear_attn.A_log": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.linear_attn.conv1d.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.linear_attn.dt_bias": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.linear_attn.in_proj_a.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.linear_attn.in_proj_b.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.linear_attn.in_proj_qkv.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.29.linear_attn.in_proj_z.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.29.linear_attn.norm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.linear_attn.out_proj.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.29.mlp.experts.down_proj": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.29.mlp.experts.gate_up_proj": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.mlp.gate.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.mlp.shared_expert.down_proj.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.mlp.shared_expert.gate_proj.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.mlp.shared_expert.up_proj.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.mlp.shared_expert_gate.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.29.post_attention_layernorm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.3.input_layernorm.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.mlp.experts.down_proj": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.mlp.experts.gate_up_proj": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.mlp.gate.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.mlp.shared_expert.down_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.mlp.shared_expert.gate_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.mlp.shared_expert.up_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.mlp.shared_expert_gate.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.post_attention_layernorm.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.self_attn.k_norm.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.self_attn.k_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.self_attn.o_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.self_attn.q_norm.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.self_attn.q_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.3.self_attn.v_proj.weight": "model-00003-of-00026.safetensors",
+    "model.language_model.layers.30.input_layernorm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.linear_attn.A_log": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.linear_attn.conv1d.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.linear_attn.dt_bias": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.linear_attn.in_proj_a.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.linear_attn.in_proj_b.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.linear_attn.in_proj_qkv.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.30.linear_attn.in_proj_z.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.30.linear_attn.norm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.linear_attn.out_proj.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.30.mlp.experts.down_proj": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.mlp.experts.gate_up_proj": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.30.mlp.gate.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.mlp.shared_expert.down_proj.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.mlp.shared_expert.gate_proj.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.mlp.shared_expert.up_proj.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.mlp.shared_expert_gate.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.30.post_attention_layernorm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.input_layernorm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.mlp.experts.down_proj": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.31.mlp.experts.gate_up_proj": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.31.mlp.gate.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.mlp.shared_expert.down_proj.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.mlp.shared_expert.gate_proj.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.mlp.shared_expert.up_proj.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.mlp.shared_expert_gate.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.post_attention_layernorm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.self_attn.k_norm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.self_attn.k_proj.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.31.self_attn.o_proj.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.31.self_attn.q_norm.weight": "model-00021-of-00026.safetensors",
+    "model.language_model.layers.31.self_attn.q_proj.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.31.self_attn.v_proj.weight": "model-00020-of-00026.safetensors",
+    "model.language_model.layers.32.input_layernorm.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.linear_attn.A_log": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.linear_attn.conv1d.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.linear_attn.dt_bias": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.linear_attn.in_proj_a.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.linear_attn.in_proj_b.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.linear_attn.in_proj_qkv.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.linear_attn.in_proj_z.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.linear_attn.norm.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.linear_attn.out_proj.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.mlp.experts.down_proj": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.mlp.experts.gate_up_proj": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.mlp.gate.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.mlp.shared_expert.down_proj.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.mlp.shared_expert.gate_proj.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.mlp.shared_expert.up_proj.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.mlp.shared_expert_gate.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.32.post_attention_layernorm.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.input_layernorm.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.linear_attn.A_log": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.linear_attn.conv1d.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.linear_attn.dt_bias": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.linear_attn.in_proj_a.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.linear_attn.in_proj_b.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.linear_attn.in_proj_qkv.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.linear_attn.in_proj_z.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.linear_attn.norm.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.linear_attn.out_proj.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.mlp.experts.down_proj": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.mlp.experts.gate_up_proj": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.mlp.gate.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.mlp.shared_expert.down_proj.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.mlp.shared_expert.gate_proj.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.mlp.shared_expert.up_proj.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.mlp.shared_expert_gate.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.33.post_attention_layernorm.weight": "model-00022-of-00026.safetensors",
+    "model.language_model.layers.34.input_layernorm.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.linear_attn.A_log": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.linear_attn.conv1d.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.linear_attn.dt_bias": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.linear_attn.in_proj_a.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.linear_attn.in_proj_b.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.linear_attn.in_proj_qkv.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.linear_attn.in_proj_z.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.linear_attn.norm.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.linear_attn.out_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.mlp.experts.down_proj": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.mlp.experts.gate_up_proj": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.mlp.gate.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.mlp.shared_expert.down_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.mlp.shared_expert.gate_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.mlp.shared_expert.up_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.mlp.shared_expert_gate.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.34.post_attention_layernorm.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.input_layernorm.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.mlp.experts.down_proj": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.mlp.experts.gate_up_proj": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.mlp.gate.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.mlp.shared_expert.down_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.mlp.shared_expert.gate_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.mlp.shared_expert.up_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.mlp.shared_expert_gate.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.post_attention_layernorm.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.self_attn.k_norm.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.self_attn.k_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.self_attn.o_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.self_attn.q_norm.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.self_attn.q_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.35.self_attn.v_proj.weight": "model-00023-of-00026.safetensors",
+    "model.language_model.layers.36.input_layernorm.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.linear_attn.A_log": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.linear_attn.conv1d.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.linear_attn.dt_bias": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.linear_attn.in_proj_a.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.linear_attn.in_proj_b.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.linear_attn.in_proj_qkv.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.linear_attn.in_proj_z.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.linear_attn.norm.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.linear_attn.out_proj.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.mlp.experts.down_proj": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.mlp.experts.gate_up_proj": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.mlp.gate.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.mlp.shared_expert.down_proj.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.mlp.shared_expert.gate_proj.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.mlp.shared_expert.up_proj.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.mlp.shared_expert_gate.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.36.post_attention_layernorm.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.input_layernorm.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.linear_attn.A_log": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.linear_attn.conv1d.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.linear_attn.dt_bias": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.linear_attn.in_proj_a.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.linear_attn.in_proj_b.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.linear_attn.in_proj_qkv.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.linear_attn.in_proj_z.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.linear_attn.norm.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.linear_attn.out_proj.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.mlp.experts.down_proj": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.mlp.experts.gate_up_proj": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.mlp.gate.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.mlp.shared_expert.down_proj.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.mlp.shared_expert.gate_proj.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.mlp.shared_expert.up_proj.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.mlp.shared_expert_gate.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.37.post_attention_layernorm.weight": "model-00024-of-00026.safetensors",
+    "model.language_model.layers.38.input_layernorm.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.linear_attn.A_log": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.linear_attn.conv1d.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.linear_attn.dt_bias": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.linear_attn.in_proj_a.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.linear_attn.in_proj_b.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.linear_attn.in_proj_qkv.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.linear_attn.in_proj_z.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.linear_attn.norm.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.linear_attn.out_proj.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.mlp.experts.down_proj": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.mlp.experts.gate_up_proj": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.mlp.gate.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.mlp.shared_expert.down_proj.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.mlp.shared_expert.gate_proj.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.mlp.shared_expert.up_proj.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.mlp.shared_expert_gate.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.38.post_attention_layernorm.weight": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.39.input_layernorm.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.mlp.experts.down_proj": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.mlp.experts.gate_up_proj": "model-00025-of-00026.safetensors",
+    "model.language_model.layers.39.mlp.gate.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.mlp.shared_expert.down_proj.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.mlp.shared_expert.gate_proj.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.mlp.shared_expert.up_proj.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.mlp.shared_expert_gate.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.post_attention_layernorm.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.self_attn.k_norm.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.self_attn.k_proj.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.self_attn.o_proj.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.self_attn.q_norm.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.self_attn.q_proj.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.39.self_attn.v_proj.weight": "model-00026-of-00026.safetensors",
+    "model.language_model.layers.4.input_layernorm.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.linear_attn.A_log": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.linear_attn.conv1d.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.linear_attn.dt_bias": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.linear_attn.in_proj_a.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.linear_attn.in_proj_b.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.linear_attn.in_proj_qkv.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.linear_attn.in_proj_z.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.linear_attn.norm.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.linear_attn.out_proj.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.mlp.experts.down_proj": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.mlp.experts.gate_up_proj": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.mlp.gate.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.mlp.shared_expert.down_proj.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.mlp.shared_expert.gate_proj.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.mlp.shared_expert.up_proj.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.mlp.shared_expert_gate.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.4.post_attention_layernorm.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.input_layernorm.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.linear_attn.A_log": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.linear_attn.conv1d.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.linear_attn.dt_bias": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.linear_attn.in_proj_a.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.linear_attn.in_proj_b.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.linear_attn.in_proj_qkv.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.linear_attn.in_proj_z.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.linear_attn.norm.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.linear_attn.out_proj.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.mlp.experts.down_proj": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.mlp.experts.gate_up_proj": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.mlp.gate.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.mlp.shared_expert.down_proj.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.mlp.shared_expert.gate_proj.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.mlp.shared_expert.up_proj.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.mlp.shared_expert_gate.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.5.post_attention_layernorm.weight": "model-00004-of-00026.safetensors",
+    "model.language_model.layers.6.input_layernorm.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.linear_attn.A_log": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.linear_attn.conv1d.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.linear_attn.dt_bias": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.linear_attn.in_proj_a.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.linear_attn.in_proj_b.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.linear_attn.in_proj_qkv.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.linear_attn.in_proj_z.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.linear_attn.norm.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.linear_attn.out_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.mlp.experts.down_proj": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.mlp.experts.gate_up_proj": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.mlp.gate.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.mlp.shared_expert.down_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.mlp.shared_expert.gate_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.mlp.shared_expert.up_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.mlp.shared_expert_gate.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.6.post_attention_layernorm.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.input_layernorm.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.mlp.experts.down_proj": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.mlp.experts.gate_up_proj": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.mlp.gate.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.mlp.shared_expert.down_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.mlp.shared_expert.gate_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.mlp.shared_expert.up_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.mlp.shared_expert_gate.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.post_attention_layernorm.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.self_attn.k_norm.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.self_attn.k_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.self_attn.o_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.self_attn.q_norm.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.self_attn.q_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.7.self_attn.v_proj.weight": "model-00005-of-00026.safetensors",
+    "model.language_model.layers.8.input_layernorm.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.linear_attn.A_log": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.linear_attn.conv1d.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.linear_attn.dt_bias": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.linear_attn.in_proj_a.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.linear_attn.in_proj_b.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.linear_attn.in_proj_qkv.weight": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.8.linear_attn.in_proj_z.weight": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.8.linear_attn.norm.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.linear_attn.out_proj.weight": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.8.mlp.experts.down_proj": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.8.mlp.experts.gate_up_proj": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.8.mlp.gate.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.mlp.shared_expert.down_proj.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.mlp.shared_expert.gate_proj.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.mlp.shared_expert.up_proj.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.mlp.shared_expert_gate.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.8.post_attention_layernorm.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.input_layernorm.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.linear_attn.A_log": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.linear_attn.conv1d.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.linear_attn.dt_bias": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.linear_attn.in_proj_a.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.linear_attn.in_proj_b.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.linear_attn.in_proj_qkv.weight": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.9.linear_attn.in_proj_z.weight": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.9.linear_attn.norm.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.linear_attn.out_proj.weight": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.9.mlp.experts.down_proj": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.9.mlp.experts.gate_up_proj": "model-00006-of-00026.safetensors",
+    "model.language_model.layers.9.mlp.gate.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.mlp.shared_expert.down_proj.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.mlp.shared_expert.gate_proj.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.mlp.shared_expert.up_proj.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.mlp.shared_expert_gate.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.layers.9.post_attention_layernorm.weight": "model-00007-of-00026.safetensors",
+    "model.language_model.norm.weight": "model-00026-of-00026.safetensors",
+    "model.visual.blocks.0.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.0.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.0.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.0.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.0.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.0.mlp.linear_fc1.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.0.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.0.mlp.linear_fc2.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.0.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.0.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.0.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.0.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.mlp.linear_fc1.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.1.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.mlp.linear_fc2.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.1.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.1.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.mlp.linear_fc1.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.10.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.mlp.linear_fc2.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.10.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.10.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.mlp.linear_fc1.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.11.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.mlp.linear_fc2.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.11.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.11.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.mlp.linear_fc1.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.12.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.mlp.linear_fc2.weight": "model-00001-of-00026.safetensors",
+    "model.visual.blocks.12.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.12.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.13.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.14.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.15.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.16.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.17.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.18.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.19.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.2.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.20.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.21.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.22.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.23.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.24.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.25.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.26.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.3.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.4.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.5.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.6.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.7.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.8.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.attn.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.attn.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.attn.qkv.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.attn.qkv.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.mlp.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.mlp.linear_fc1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.mlp.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.mlp.linear_fc2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.norm1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.norm1.weight": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.norm2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.blocks.9.norm2.weight": "model-00002-of-00026.safetensors",
+    "model.visual.merger.linear_fc1.bias": "model-00002-of-00026.safetensors",
+    "model.visual.merger.linear_fc1.weight": "model-00001-of-00026.safetensors",
+    "model.visual.merger.linear_fc2.bias": "model-00002-of-00026.safetensors",
+    "model.visual.merger.linear_fc2.weight": "model-00001-of-00026.safetensors",
+    "model.visual.merger.norm.bias": "model-00002-of-00026.safetensors",
+    "model.visual.merger.norm.weight": "model-00002-of-00026.safetensors",
+    "model.visual.patch_embed.proj.bias": "model-00002-of-00026.safetensors",
+    "model.visual.patch_embed.proj.weight": "model-00002-of-00026.safetensors",
+    "model.visual.pos_embed.weight": "model-00002-of-00026.safetensors",
+    "mtp.fc.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.input_layernorm.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.mlp.experts.down_proj": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.mlp.experts.gate_up_proj": "model-00025-of-00026.safetensors",
+    "mtp.layers.0.mlp.gate.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.mlp.shared_expert.down_proj.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.mlp.shared_expert.gate_proj.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.mlp.shared_expert.up_proj.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.mlp.shared_expert_gate.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.post_attention_layernorm.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.self_attn.k_norm.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.self_attn.k_proj.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.self_attn.o_proj.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.self_attn.q_norm.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.self_attn.q_proj.weight": "model-00026-of-00026.safetensors",
+    "mtp.layers.0.self_attn.v_proj.weight": "model-00026-of-00026.safetensors",
+    "mtp.norm.weight": "model-00026-of-00026.safetensors",
+    "mtp.pre_fc_norm_embedding.weight": "model-00026-of-00026.safetensors",
+    "mtp.pre_fc_norm_hidden.weight": "model-00026-of-00026.safetensors"
+  }
+}

preprocessor_config.json

@@ -0,0 +1,21 @@
+{
+    "size": {
+        "longest_edge": 16777216,
+        "shortest_edge": 65536
+    },
+    "patch_size": 16,
+    "temporal_patch_size": 2,
+    "merge_size": 2,
+    "image_mean": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "image_std": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "processor_class": "Qwen3VLProcessor",
+    "image_processor_type": "Qwen2VLImageProcessorFast"
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f9e4d4901a92b997e463c1f46055088b6cca5ca61a6522d1b9f64c4bb81cb42
+size 12807982

tokenizer_config.json

@@ -0,0 +1,305 @@
+{
+    "add_prefix_space": false,
+    "added_tokens_decoder": {
+        "248044": {
+            "content": "<|endoftext|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248045": {
+            "content": "<|im_start|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248046": {
+            "content": "<|im_end|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248047": {
+            "content": "<|object_ref_start|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248048": {
+            "content": "<|object_ref_end|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248049": {
+            "content": "<|box_start|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248050": {
+            "content": "<|box_end|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248051": {
+            "content": "<|quad_start|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248052": {
+            "content": "<|quad_end|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248053": {
+            "content": "<|vision_start|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248054": {
+            "content": "<|vision_end|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248055": {
+            "content": "<|vision_pad|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248056": {
+            "content": "<|image_pad|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248057": {
+            "content": "<|video_pad|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248058": {
+            "content": "<tool_call>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248059": {
+            "content": "</tool_call>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248060": {
+            "content": "<|fim_prefix|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248061": {
+            "content": "<|fim_middle|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248062": {
+            "content": "<|fim_suffix|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248063": {
+            "content": "<|fim_pad|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248064": {
+            "content": "<|repo_name|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248065": {
+            "content": "<|file_sep|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248066": {
+            "content": "<tool_response>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248067": {
+            "content": "</tool_response>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248068": {
+            "content": "<think>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248069": {
+            "content": "</think>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": false
+        },
+        "248070": {
+            "content": "<|audio_start|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248071": {
+            "content": "<|audio_end|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248072": {
+            "content": "<tts_pad>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248073": {
+            "content": "<tts_text_bos>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248074": {
+            "content": "<tts_text_eod>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248075": {
+            "content": "<tts_text_bos_single>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        },
+        "248076": {
+            "content": "<|audio_pad|>",
+            "lstrip": false,
+            "normalized": false,
+            "rstrip": false,
+            "single_word": false,
+            "special": true
+        }
+    },
+    "additional_special_tokens": [
+        "<|im_start|>",
+        "<|im_end|>",
+        "<|object_ref_start|>",
+        "<|object_ref_end|>",
+        "<|box_start|>",
+        "<|box_end|>",
+        "<|quad_start|>",
+        "<|quad_end|>",
+        "<|vision_start|>",
+        "<|vision_end|>",
+        "<|vision_pad|>",
+        "<|image_pad|>",
+        "<|video_pad|>"
+    ],
+    "bos_token": null,
+    "chat_template": "{%- set image_count = namespace(value=0) %}\n{%- set video_count = namespace(value=0) %}\n{%- macro render_content(content, do_vision_count, is_system_content=false) %}\n    {%- if content is string %}\n        {{- content }}\n    {%- elif content is iterable and content is not mapping %}\n        {%- for item in content %}\n            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}\n                {%- if is_system_content %}\n                    {{- raise_exception('System message cannot contain images.') }}\n                {%- endif %}\n                {%- if do_vision_count %}\n                    {%- set image_count.value = image_count.value + 1 %}\n                {%- endif %}\n                {%- if add_vision_id %}\n                    {{- 'Picture ' ~ image_count.value ~ ': ' }}\n                {%- endif %}\n                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}\n            {%- elif 'video' in item or item.type == 'video' %}\n                {%- if is_system_content %}\n                    {{- raise_exception('System message cannot contain videos.') }}\n                {%- endif %}\n                {%- if do_vision_count %}\n                    {%- set video_count.value = video_count.value + 1 %}\n                {%- endif %}\n                {%- if add_vision_id %}\n                    {{- 'Video ' ~ video_count.value ~ ': ' }}\n                {%- endif %}\n                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}\n            {%- elif 'text' in item %}\n                {{- item.text }}\n            {%- else %}\n                {{- raise_exception('Unexpected item type in content.') }}\n            {%- endif %}\n        {%- endfor %}\n    {%- elif content is none or content is undefined %}\n        {{- '' }}\n    {%- else %}\n        {{- raise_exception('Unexpected content type.') }}\n    {%- endif %}\n{%- endmacro %}\n{%- if not messages %}\n    {{- raise_exception('No messages provided.') }}\n{%- endif %}\n{%- if tools and tools is iterable and tools is not mapping %}\n    {{- '<|im_start|>system\\n' }}\n    {{- \"# Tools\\n\\nYou have access to the following functions:\\n\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\" }}\n    {{- '\\n\\nIf you choose to call a function ONLY reply in the following format with NO suffix:\\n\\n<tool_call>\\n<function=example_function_name>\\n<parameter=example_parameter_1>\\nvalue_1\\n</parameter>\\n<parameter=example_parameter_2>\\nThis is the value for the second parameter\\nthat can span\\nmultiple lines\\n</parameter>\\n</function>\\n</tool_call>\\n\\n<IMPORTANT>\\nReminder:\\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\\n- Required parameters MUST be specified\\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\\n</IMPORTANT>' }}\n    {%- if messages[0].role == 'system' %}\n        {%- set content = render_content(messages[0].content, false, true)|trim %}\n        {%- if content %}\n            {{- '\\n\\n' + content }}\n        {%- endif %}\n    {%- endif %}\n    {{- '<|im_end|>\\n' }}\n{%- else %}\n    {%- if messages[0].role == 'system' %}\n        {%- set content = render_content(messages[0].content, false, true)|trim %}\n        {{- '<|im_start|>system\\n' + content + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" %}\n        {%- set content = render_content(message.content, false)|trim %}\n        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}\n            {%- set ns.multi_step_tool = false %}\n            {%- set ns.last_query_index = index %}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if ns.multi_step_tool %}\n    {{- raise_exception('No user query found in messages.') }}\n{%- endif %}\n{%- for message in messages %}\n    {%- set content = render_content(message.content, true)|trim %}\n    {%- if message.role == \"system\" %}\n        {%- if not loop.first %}\n            {{- raise_exception('System message must be at the beginning.') }}\n        {%- endif %}\n    {%- elif message.role == \"user\" %}\n        {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is string %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in content %}\n                {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n                {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- set reasoning_content = reasoning_content|trim %}\n        {%- if (preserve_thinking is defined and preserve_thinking is true) or (loop.index0 > ns.last_query_index) %}\n            {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content + '\\n</think>\\n\\n' + content }}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if tool_call.function is defined %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {%- if loop.first %}\n                    {%- if content|trim %}\n                        {{- '\\n\\n<tool_call>\\n<function=' + tool_call.name + '>\\n' }}\n                    {%- else %}\n                        {{- '<tool_call>\\n<function=' + tool_call.name + '>\\n' }}\n                    {%- endif %}\n                {%- else %}\n                    {{- '\\n<tool_call>\\n<function=' + tool_call.name + '>\\n' }}\n                {%- endif %}\n                {%- if tool_call.arguments is defined %}\n                    {%- for args_name, args_value in tool_call.arguments|items %}\n                        {{- '<parameter=' + args_name + '>\\n' }}\n                        {%- set args_value = args_value | string if args_value is string else args_value | tojson | safe %}\n                        {{- args_value }}\n                        {{- '\\n</parameter>\\n' }}\n                    {%- endfor %}\n                {%- endif %}\n                {{- '</function>\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.previtem and loop.previtem.role != \"tool\" %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- content }}\n        {{- '\\n</tool_response>' }}\n        {%- if not loop.last and loop.nextitem.role != \"tool\" %}\n            {{- '<|im_end|>\\n' }}\n        {%- elif loop.last %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- else %}\n        {{- raise_exception('Unexpected message role.') }}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- else %}\n        {{- '<think>\\n' }}\n    {%- endif %}\n{%- endif %}",
+    "clean_up_tokenization_spaces": false,
+    "eos_token": "<|im_end|>",
+    "errors": "replace",
+    "model_max_length": 262144,
+    "pad_token": "<|endoftext|>",
+    "split_special_tokens": false,
+    "tokenizer_class": "Qwen2Tokenizer",
+    "unk_token": null,
+    "add_bos_token": false,
+    "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+    "extra_special_tokens": {
+        "audio_bos_token": "<|audio_start|>",
+        "audio_eos_token": "<|audio_end|>",
+        "audio_token": "<|audio_pad|>",
+        "image_token": "<|image_pad|>",
+        "video_token": "<|video_pad|>",
+        "vision_bos_token": "<|vision_start|>",
+        "vision_eos_token": "<|vision_end|>"
+    }
+}

video_preprocessor_config.json

@@ -0,0 +1,21 @@
+{
+    "size": {
+        "longest_edge": 25165824,
+        "shortest_edge": 4096
+    },
+    "patch_size": 16,
+    "temporal_patch_size": 2,
+    "merge_size": 2,
+    "image_mean": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "image_std": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "processor_class": "Qwen3VLProcessor",
+    "video_processor_type": "Qwen3VLVideoProcessor"
+}