Upload folder using huggingface_hub

.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

README.md

@@ -0,0 +1,152 @@
+---
+language:
+  - en
+  - zh
+library_name: transformers
+license: mit
+pipeline_tag: text-generation
+---
+
+# GLM-5
+
+<div align="center">
+<img src=https://raw.githubusercontent.com/zai-org/GLM-5/refs/heads/main/resources/logo.svg width="15%"/>
+</div>
+<p align="center">
+    👋 Join our <a href="https://raw.githubusercontent.com/zai-org/GLM-5/refs/heads/main/resources/wechat.png" target="_blank">WeChat</a> or <a href="https://discord.gg/QR7SARHRxK" target="_blank">Discord</a> community.
+    <br>
+    📖 Check out the GLM-5 <a href="https://z.ai/blog/glm-5" target="_blank">technical blog</a>.
+    <br>
+    📍 Use GLM-5 API services on <a href="https://docs.z.ai/guides/llm/glm-5">Z.ai API Platform. </a>
+    <br>
+    👉 One click to <a href="https://chat.z.ai">GLM-5</a>.
+</p>
+
+## Introduction
+
+We are launching GLM-5, targeting complex systems engineering and long-horizon agentic tasks. Scaling is still one of the most important ways to improve the intelligence efficiency of Artificial General Intelligence (AGI). Compared to GLM-4.5, GLM-5 scales from 355B parameters (32B active) to 744B parameters (40B active), and increases pre-training data from 23T to 28.5T tokens. GLM-5 also integrates DeepSeek Sparse Attention (DSA), largely reducing deployment cost while preserving long-context capacity.
+
+Reinforcement learning aims to bridge the gap between competence and excellence in pre-trained models. However, deploying it at scale for LLMs is a challenge due to the RL training inefficiency. To this end, we developed [slime](https://github.com/THUDM/slime), a novel **asynchronous RL infrastructure** that substantially improves training throughput and efficiency, enabling more fine-grained post-training iterations. With advances in both pre-training and post-training, GLM-5 delivers significant improvement compared to GLM-4.7 across a wide range of academic benchmarks and achieves best-in-class performance among all open-source models in the world on reasoning, coding, and agentic tasks,  closing the gap with frontier models.
+
+## Benchmark
+
+|                                  | GLM-5                  | GLM-4.7   | DeepSeek-V3.2 | Kimi K2.5 | Claude Opus 4.5 | Gemini 3 Pro | GPT-5.2 (xhigh) |
+| -------------------------------- | ---------------------- | --------- | ------------- |-----------| --------------- | ------------ | --------------- |
+| HLE                              | 30.5                   | 24.8      | 25.1          | 31.5      | 28.4            | 37.2         | 35.4            |
+| HLE (w/ Tools)                   | 50.4                   | 42.8      | 40.8          | 51.8      | 43.4*           | 45.8*        | 45.5*           |
+| AIME 2026 I                      | 92.7                   | 92.9      | 92.7          | 92.5      | 93.3            | 90.6         | -               |
+| HMMT Nov. 2025                   | 96.9                   | 93.5      | 90.2          | 91.1      | 91.7            | 93.0         | 97.1            |
+| IMOAnswerBench                   | 82.5                   | 82.0      | 78.3          | 81.8      | 78.5            | 83.3         | 86.3            |
+| GPQA-Diamond                     | 86.0                   | 85.7      | 82.4          | 87.6      | 87.0            | 91.9         | 92.4            |
+| SWE-bench Verified               | 77.8                   | 73.8      | 73.1          | 76.8      | 80.9            | 76.2         | 80.0            |
+| SWE-bench Multilingual           | 73.3                   | 66.7      | 70.2          | 73.0      | 77.5            | 65.0         | 72.0            |
+| Terminal-Bench 2.0 (Terminus 2)  | 56.2 / 60.7 † | 41.0      | 39.3          | 50.8      | 59.3            | 54.2         | 54.0            |
+| Terminal-Bench 2.0 (Claude Code) | 56.2 / 61.1 †  | 32.8      | 46.4          | -         | 57.9            | -            | -               |
+| CyberGym                         | 43.2                   | 23.5      | 17.3          | 41.3      | 50.6            | 39.9         | -               |
+| BrowseComp                       | 62.0                   | 52.0      | 51.4          | 60.6      | 37.0            | 37.8         | -               |
+| BrowseComp (w/ Context Manage)   | 75.9                   | 67.5      | 67.6          | 74.9      | 67.8            | 59.2         | 65.8            |
+| BrowseComp-Zh                    | 72.7                   | 66.6      | 65.0          | 62.3      | 62.4            | 66.8         | 76.1            |
+| τ²-Bench                         | 89.7                   | 87.4      | 85.3          | 80.2      | 91.6            | 90.7         | 85.5            |
+| MCP-Atlas (Public Set)           | 67.8                   | 52.0      | 62.2          | 63.8      | 65.2            | 66.6         | 68.0            |
+| Tool-Decathlon                   | 38.0                   | 23.8      | 35.2          | 27.8      | 43.5            | 36.4         | 46.3            |
+| Vending Bench 2                  | $4,432.12              | $2,376.82 | $1,034.00     | $1,198.46 | $4,967.06       | $5,478.16    | $3,591.33       |
+
+> *: refers to their scores of full set.
+> 
+> †: A verified version of Terminal-Bench 2.0 that fixes some ambiguous instructions.
+See footnote for more evaluation details.
+
+### Footnote
+
+* **Humanity’s Last Exam (HLE) & other reasoning tasks**: We evaluate with a maximum generation length of 131,072 tokens (`temperature=1.0, top_p=0.95, max_new_tokens=131072`). By default, we report the text-only subset; results marked with * are from the full set. We use GPT-5.2 (medium) as the judge model. For HLE-with-tools, we use a maximum context length of 202,752 tokens.
+* **SWE-bench & SWE-bench Multilingual**: We run the SWE-bench suite with OpenHands using a tailored instruction prompt. Settings: `temperature=0.7, top_p=0.95, max_new_tokens=16384`, with a 200K context window.
+* **BrowserComp**: Without context management, we retain details from the most recent 5 turns. With context management, we use the same discard-all strategy as DeepSeek-v3.2 and Kimi K2.5.
+* **Terminal-Bench 2.0 (Terminus 2)**: We evaluate with the Terminus framework using `timeout=2h, temperature=0.7, top_p=1.0, max_new_tokens=8192`, with a 128K context window. Resource limits are capped at 16 CPUs and 32 GB RAM.
+* **Terminal-Bench 2.0 (Claude Code)**: We evaluate in Claude Code 2.1.14 (think mode, default effort) with `temperature=1.0, top_p=0.95, max_new_tokens=65536`. We remove wall-clock time limits due to generation speed, while preserving per-task CPU and memory constraints. Scores are averaged over 5 runs. We fix environment issues introduced by Claude Code and also report results on a verified Terminal-Bench 2.0 dataset that resolves ambiguous instructions (see: [https://huggingface.co/datasets/zai-org/terminal-bench-2-verified](https://huggingface.co/datasets/zai-org/terminal-bench-2-verified)).
+* **CyberGym**: We evaluate in Claude Code 2.1.18 (think mode, no web tools) with (`temperature=1.0, top_p=1.0, max_new_tokens=32000`) and a 250-minute timeout per task. Results are single-run Pass@1 over 1,507 tasks.
+* **MCP-Atlas**: All models are evaluated in think mode on the 500-task public subset with a 10-minute timeout per task. We use Gemini 3 Pro as the judge model.
+* **τ²-bench**: We add a small prompt adjustment in Retail and Telecom to avoid failures caused by premature user termination. For Airline, we apply the domain fixes proposed in the Claude Opus 4.5 system card.
+* **Vending Bench 2**: Runs are conducted independently by [Andon Labs](https://andonlabs.com/evals/vending-bench-2).
+
+
+## Serve GLM-5 Locally
+
+### Prepare environment
+
+vLLM, SGLang, KTransformers, and xLLM all support local deployment of GLM-5. A simple deployment guide is provided here.
+
++ vLLM
+
+    Using Docker as:
+
+    ```shell
+    docker pull vllm/vllm-openai:nightly 
+    ```
+
+    or using pip:
+
+    ```shell
+    pip install -U vllm --pre --index-url https://pypi.org/simple --extra-index-url https://wheels.vllm.ai/nightly
+    ```
+
+    then upgrade transformers:
+
+    ```
+    pip install git+https://github.com/huggingface/transformers.git
+    ```
+
++ SGLang
+
+    Using Docker as:
+    ```bash
+    docker pull lmsysorg/sglang:glm5-hopper # For Hopper GPU
+    docker pull lmsysorg/sglang:glm5-blackwell # For Blackwell GPU
+    ```
+
+### Deploy
+
++ vLLM
+
+    ```shell
+    vllm serve zai-org/GLM-5-FP8 \
+         --tensor-parallel-size 8 \
+         --gpu-memory-utilization 0.85 \
+         --speculative-config.method mtp \
+         --speculative-config.num_speculative_tokens 1 \
+         --tool-call-parser glm47 \
+         --reasoning-parser glm45 \
+         --enable-auto-tool-choice \
+         --served-model-name glm-5-fp8
+    ```
+
+    Check the [recipes](https://github.com/vllm-project/recipes/blob/main/GLM/GLM5.md) for more details.
+
++ SGLang
+
+    ```shell
+    python3 -m sglang.launch_server \
+      --model-path zai-org/GLM-5-FP8 \
+      --tp-size 8 \
+      --tool-call-parser glm47  \
+      --reasoning-parser glm45 \
+      --speculative-algorithm EAGLE \
+      --speculative-num-steps 3 \
+      --speculative-eagle-topk 1 \
+      --speculative-num-draft-tokens 4 \
+      --mem-fraction-static 0.85 \
+      --served-model-name glm-5-fp8
+    ```
+  
+    Check the [sglang cookbook](https://cookbook.sglang.io/autoregressive/GLM/GLM-5) for more details.
+
++ xLLM and other Ascend NPU
+
+    Please check the deployment guide [here](https://github.com/zai-org/GLM-5/blob/main/example/ascend.md).
+
++ KTransformers
+
+    Please check the deployment guide [here](https://github.com/kvcache-ai/ktransformers/blob/main/doc/en/kt-kernel/GLM-5-Tutorial.md).
+
+## Citation
+
+Our technical report is coming soon.

chat_template.jinja

@@ -0,0 +1,86 @@
+[gMASK]<sop>
+{%- if tools -%}
+<|system|>
+# Tools
+
+You may call one or more functions to assist with the user query.
+
+You are provided with function signatures within <tools></tools> XML tags:
+<tools>
+{% for tool in tools %}
+{{ tool | tojson(ensure_ascii=False) }}
+{% endfor %}
+</tools>
+
+For each function call, output the function name and arguments within the following XML format:
+<tool_call>{function-name}<arg_key>{arg-key-1}</arg_key><arg_value>{arg-value-1}</arg_value><arg_key>{arg-key-2}</arg_key><arg_value>{arg-value-2}</arg_value>...</tool_call>{%- endif -%}
+{%- macro visible_text(content) -%}
+    {%- if content is string -%}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping -%}
+        {%- for item in content -%}
+            {%- if item is mapping and item.type == 'text' -%}
+                {{- item.text }}
+            {%- elif item is string -%}
+                {{- item }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{- content }}
+    {%- endif -%}
+{%- endmacro -%}
+{%- set ns = namespace(last_user_index=-1) %}
+{%- for m in messages %}
+    {%- if m.role == 'user' %}
+        {% set ns.last_user_index = loop.index0 -%}
+    {%- endif %}
+{%- endfor %}
+{% for m in messages %}
+{%- if m.role == 'user' -%}<|user|>{{ visible_text(m.content) }}
+{%- elif m.role == 'assistant' -%}
+<|assistant|>
+{%- set reasoning_content = '' %}
+{%- set content = visible_text(m.content) %}
+{%- if m.reasoning_content is string %}
+    {%- set reasoning_content = m.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 %}
+{%- if ((clear_thinking is defined and not clear_thinking) or loop.index0 > ns.last_user_index) and reasoning_content -%}
+{{ '<think>' + reasoning_content.strip() +  '</think>'}}
+{%- else -%}
+{{ '</think>' }}
+{%- endif -%}
+{%- if content.strip() -%}
+{{ content.strip() }}
+{%- endif -%}
+{% if m.tool_calls %}
+{% for tc in m.tool_calls %}
+{%- if tc.function %}
+    {%- set tc = tc.function %}
+{%- endif %}
+{{- '<tool_call>' + tc.name -}}
+{% set _args = tc.arguments %}{% for k, v in _args.items() %}<arg_key>{{ k }}</arg_key><arg_value>{{ v | tojson(ensure_ascii=False) if v is not string else v }}</arg_value>{% endfor %}</tool_call>{% endfor %}
+{% endif %}
+{%- elif m.role == 'tool' -%}
+{%- if m.content is string -%}
+{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+    {{- '<|observation|>' }}
+{%- endif %}
+{{- '<tool_response>' }}
+{{- m.content }}
+{{- '</tool_response>' }}
+{%- else -%}
+<|observation|>{% for tr in m.content %}
+<tool_response>{{ tr.output if tr.output is defined else tr }}</tool_response>{% endfor -%}
+{% endif -%}
+{%- elif m.role == 'system' -%}
+<|system|>{{ visible_text(m.content) }}
+{%- endif -%}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    <|assistant|>{{- '</think>' if (enable_thinking is defined and not enable_thinking) else '<think>' -}}
+{%- endif -%}

config.json

@@ -0,0 +1,63 @@
+{
+  "architectures": [
+    "DeepseekV32ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "dtype": "bfloat16",
+  "eos_token_id": [
+    154820,
+    154827,
+    154829
+  ],
+  "ep_size": 1,
+  "first_k_dense_replace": 3,
+  "hidden_act": "silu",
+  "head_dim": 64,
+  "hidden_size": 6144,
+  "index_head_dim": 128,
+  "index_n_heads": 32,
+  "index_topk": 2048,
+  "indexer_rope_interleave": true,
+  "initializer_range": 0.02,
+  "intermediate_size": 12288,
+  "kv_lora_rank": 512,
+  "max_position_embeddings": 202752,
+  "moe_intermediate_size": 2048,
+  "moe_layer_freq": 1,
+  "model_type": "deepseek_v32",
+  "n_group": 1,
+  "n_routed_experts": 256,
+  "n_shared_experts": 1,
+  "norm_topk_prob": true,
+  "num_attention_heads": 64,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 4,
+  "num_key_value_heads": 64,
+  "num_nextn_predict_layers": 1,
+  "pad_token_id": 154820,
+  "pretraining_tp": 1,
+  "q_lora_rank": 2048,
+  "qk_head_dim": 256,
+  "qk_nope_head_dim": 192,
+  "qk_rope_head_dim": 64,
+  "rms_norm_eps": 1e-05,
+  "rope_interleave": true,
+  "rope_parameters": {
+    "rope_theta": 1000000,
+    "rope_type": "default"
+  },
+  "routed_scaling_factor": 2.5,
+  "scoring_func": "sigmoid",
+  "tie_word_embeddings": false,
+  "topk_group": 1,
+  "topk_method": "noaux_tc",
+  "transformers_version": "5.0.2.dev0",
+  "use_cache": true,
+  "v_head_dim": 256,
+  "vocab_size": 154880,
+  "auto_map": {
+    "AutoConfig": "configuration_deepseek_v32.DeepseekV32Config",
+    "AutoModelForCausalLM": "modeling_deepseek_v32.DeepseekV32ForCausalLM"
+  }
+}

configuration_deepseek_v32.py

@@ -0,0 +1,61 @@
+# coding=utf-8
+# Copyright 2025 bzantium and the HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on the DeepSeekV3 implementations from the DeepSeek AI team. (https://huggingface.co/deepseek-ai/DeepSeek-V3)
+
+# 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.
+"""DeepSeekV3.2 model configuration"""
+
+from typing import Optional
+
+from transformers.models.deepseek_v3.configuration_deepseek_v3 import DeepseekV3Config
+
+
+DEEPSEEK_V32_PRETRAINED_CONFIG_ARCHIVE_MAP = {}
+
+
+class DeepseekV32Config(DeepseekV3Config):
+    r"""
+    This is the configuration class to store the configuration of a [`DeepseekV32Model`]. It is used to instantiate a DeepSeek
+    V3.2 model according to the specified arguments, defining the model architecture. 
+    
+    DeepSeek V3.2 extends DeepSeek V3 with native sparse attention mechanism using an indexer for efficient
+    attention computation on long sequences.
+    Configuration objects inherit from [`DeepseekV3Config`] and can be used to control the model outputs. Read the
+    documentation from [`PreTrainedConfig`] for more information.
+    Args:
+        index_topk (`int`, *optional*, defaults to 2048):
+            Number of top-k tokens to select for sparse attention. This enables the native sparse attention
+            mechanism in DeepSeek V3.2.
+        **kwargs:
+            All other arguments from DeepseekV3Config.
+    ```python
+    >>> from transformers import DeepseekV32Model, DeepseekV32Config
+    >>> # Initializing a Deepseek-V3.2 style configuration
+    >>> configuration = DeepseekV32Config()
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "deepseek_v32"
+
+    def __init__(
+        self,
+        index_topk: Optional[int] = 2048,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        self.index_topk = index_topk
+
+
+__all__ = ["DeepseekV32Config"]

generation_config.json

@@ -0,0 +1,12 @@
+{
+  "_from_model_config": true,
+  "eos_token_id": [
+    154820,
+    154827,
+    154829
+  ],
+  "pad_token_id": 154820,
+  "temperature": 1.0,
+  "top_p": 0.95,
+  "transformers_version": "5.0.2.dev0"
+}

model-00001-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:198ef923a7ca4effc5ead8ebf799fee10beb8ce081352fb099636f805d1deda9
+size 5342821416

model-00002-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd6c14c9bf4a9695408f1c8e62c03287c532a3759a7cd84952a13073d9dbe313
+size 67108984

model-00038-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c648ea022d4fd7171598777d4e6d9b52c548eae99b1cdad74ddf7c74a7c6713d
+size 709524696

model-00039-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2709fa9fee5c27e8e8a4093db5eaa577e8c612e442a4691054e8aada6f19c67b
+size 92274928

model-00075-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61269b6447c93de8cbcb9350251a8443aaad6e818f758855e76f0a1c1b4fdb1a
+size 679492952

model-00076-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:807bf5e2cd4beaa0594d692cb8939a5be25a404cdb188c23e9697791115b1e77
+size 5360347104

model-00077-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:316499d677b61f3748fe237624c07ca84342fd79b84beb8dc7fe4ac64b2918c0
+size 5360347104

model-00078-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b7faf14ea9f2eebd31f1c600031adece58e82e42cec6ad1ef57cc9a80ae4133
+size 5360346968

model-00079-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d52dfd2e3d0e8c8a09c682bded9597256640f2c27549ff8b2d1b5f727df7034f
+size 2994373544

model-00282-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee2233ec70c39f22fa82951102c588c2fc0fb24802dad52fb14b285731eb118b
+size 12376

model.safetensors.index.json

@@ -0,0 +1,848 @@
+{
+  "metadata": {
+    "total_size": 25966545920
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00001-of-00282.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.indexer.k_norm.bias": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.indexer.k_norm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.indexer.weights_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.indexer.wk.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.indexer.wq_b.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.kv_a_layernorm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.kv_a_proj_with_mqa.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.kv_b_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.q_a_layernorm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.q_a_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.0.self_attn.q_b_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.indexer.k_norm.bias": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.indexer.k_norm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.indexer.weights_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.indexer.wk.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.indexer.wq_b.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.kv_a_layernorm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.kv_a_proj_with_mqa.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.kv_b_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.q_a_layernorm.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.q_a_proj.weight": "model-00001-of-00282.safetensors",
+    "model.layers.1.self_attn.q_b_proj.weight": "model-00002-of-00282.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.indexer.k_norm.bias": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.indexer.k_norm.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.indexer.weights_proj.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.indexer.wk.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.indexer.wq_b.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.kv_a_layernorm.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.kv_a_proj_with_mqa.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.kv_b_proj.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.q_a_layernorm.weight": "model-00038-of-00282.safetensors",
+    "model.layers.2.self_attn.q_a_proj.weight": "model-00039-of-00282.safetensors",
+    "model.layers.2.self_attn.q_b_proj.weight": "model-00039-of-00282.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.0.down_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.0.gate_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.0.up_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.1.down_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.1.gate_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.1.up_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.10.down_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.10.gate_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.10.up_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.100.down_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.100.gate_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.100.up_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.101.down_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.101.gate_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.101.up_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.102.down_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.102.gate_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.102.up_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.103.down_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.103.gate_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.103.up_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.104.down_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.104.gate_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.104.up_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.105.down_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.105.gate_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.105.up_proj.weight": "model-00075-of-00282.safetensors",
+    "model.layers.3.mlp.experts.106.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.106.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.106.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.107.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.107.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.107.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.108.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.108.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.108.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.109.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.109.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.109.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.11.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.11.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.11.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.110.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.110.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.110.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.111.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.111.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.111.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.112.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.112.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.112.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.113.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.113.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.113.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.114.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.114.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.114.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.115.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.115.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.115.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.116.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.116.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.116.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.117.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.117.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.117.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.118.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.118.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.118.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.119.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.119.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.119.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.12.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.12.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.12.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.120.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.120.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.120.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.121.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.121.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.121.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.122.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.122.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.122.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.123.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.123.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.123.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.124.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.124.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.124.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.125.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.125.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.125.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.126.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.126.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.126.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.127.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.127.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.127.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.128.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.128.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.128.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.129.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.129.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.129.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.13.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.13.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.13.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.130.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.130.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.130.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.131.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.131.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.131.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.132.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.132.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.132.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.133.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.133.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.133.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.134.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.134.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.134.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.135.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.135.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.135.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.136.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.136.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.136.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.137.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.137.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.137.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.138.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.138.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.138.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.139.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.139.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.139.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.14.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.14.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.14.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.140.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.140.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.140.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.141.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.141.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.141.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.142.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.142.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.142.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.143.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.143.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.143.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.144.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.144.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.144.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.145.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.145.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.145.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.146.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.146.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.146.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.147.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.147.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.147.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.148.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.148.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.148.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.149.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.149.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.149.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.15.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.15.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.15.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.150.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.150.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.150.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.151.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.151.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.151.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.152.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.152.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.152.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.153.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.153.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.153.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.154.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.154.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.154.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.155.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.155.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.155.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.156.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.156.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.156.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.157.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.157.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.157.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.158.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.158.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.158.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.159.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.159.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.159.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.16.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.16.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.16.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.160.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.160.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.160.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.161.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.161.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.161.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.162.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.162.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.162.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.163.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.163.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.163.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.164.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.164.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.164.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.165.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.165.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.165.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.166.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.166.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.166.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.167.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.167.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.167.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.168.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.168.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.168.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.169.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.169.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.169.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.17.down_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.17.gate_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.17.up_proj.weight": "model-00076-of-00282.safetensors",
+    "model.layers.3.mlp.experts.170.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.170.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.170.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.171.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.171.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.171.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.172.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.172.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.172.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.173.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.173.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.173.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.174.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.174.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.174.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.175.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.175.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.175.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.176.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.176.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.176.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.177.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.177.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.177.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.178.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.178.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.178.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.179.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.179.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.179.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.18.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.18.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.18.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.180.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.180.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.180.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.181.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.181.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.181.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.182.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.182.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.182.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.183.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.183.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.183.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.184.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.184.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.184.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.185.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.185.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.185.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.186.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.186.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.186.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.187.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.187.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.187.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.188.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.188.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.188.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.189.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.189.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.189.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.19.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.19.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.19.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.190.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.190.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.190.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.191.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.191.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.191.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.192.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.192.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.192.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.193.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.193.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.193.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.194.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.194.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.194.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.195.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.195.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.195.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.196.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.196.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.196.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.197.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.197.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.197.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.198.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.198.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.198.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.199.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.199.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.199.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.2.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.2.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.2.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.20.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.20.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.20.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.200.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.200.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.200.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.201.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.201.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.201.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.202.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.202.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.202.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.203.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.203.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.203.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.204.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.204.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.204.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.205.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.205.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.205.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.206.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.206.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.206.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.207.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.207.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.207.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.208.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.208.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.208.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.209.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.209.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.209.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.21.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.21.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.21.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.210.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.210.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.210.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.211.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.211.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.211.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.212.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.212.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.212.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.213.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.213.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.213.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.214.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.214.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.214.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.215.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.215.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.215.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.216.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.216.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.216.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.217.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.217.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.217.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.218.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.218.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.218.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.219.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.219.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.219.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.22.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.22.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.22.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.220.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.220.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.220.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.221.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.221.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.221.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.222.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.222.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.222.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.223.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.223.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.223.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.224.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.224.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.224.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.225.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.225.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.225.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.226.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.226.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.226.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.227.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.227.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.227.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.228.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.228.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.228.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.229.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.229.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.229.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.23.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.23.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.23.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.230.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.230.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.230.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.231.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.231.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.231.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.232.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.232.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.232.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.233.down_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.233.gate_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.233.up_proj.weight": "model-00077-of-00282.safetensors",
+    "model.layers.3.mlp.experts.234.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.234.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.234.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.235.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.235.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.235.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.236.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.236.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.236.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.237.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.237.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.237.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.238.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.238.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.238.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.239.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.239.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.239.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.24.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.24.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.24.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.240.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.240.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.240.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.241.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.241.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.241.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.242.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.242.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.242.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.243.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.243.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.243.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.244.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.244.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.244.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.245.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.245.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.245.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.246.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.246.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.246.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.247.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.247.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.247.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.248.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.248.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.248.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.249.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.249.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.249.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.25.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.25.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.25.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.250.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.250.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.250.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.251.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.251.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.251.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.252.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.252.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.252.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.253.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.253.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.253.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.254.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.254.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.254.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.255.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.255.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.255.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.26.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.26.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.26.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.27.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.27.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.27.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.28.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.28.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.28.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.29.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.29.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.29.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.3.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.3.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.3.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.30.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.30.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.30.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.31.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.31.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.31.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.32.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.32.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.32.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.33.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.33.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.33.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.34.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.34.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.34.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.35.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.35.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.35.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.36.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.36.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.36.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.37.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.37.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.37.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.38.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.38.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.38.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.39.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.39.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.39.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.4.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.4.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.4.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.40.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.40.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.40.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.41.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.41.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.41.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.42.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.42.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.42.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.43.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.43.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.43.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.44.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.44.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.44.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.45.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.45.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.45.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.46.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.46.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.46.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.47.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.47.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.47.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.48.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.48.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.48.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.49.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.49.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.49.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.5.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.5.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.5.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.50.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.50.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.50.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.51.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.51.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.51.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.52.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.52.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.52.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.53.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.53.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.53.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.54.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.54.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.54.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.55.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.55.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.55.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.56.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.56.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.56.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.57.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.57.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.57.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.58.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.58.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.58.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.59.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.59.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.59.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.6.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.6.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.6.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.60.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.60.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.60.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.61.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.61.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.61.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.62.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.62.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.62.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.63.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.63.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.63.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.64.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.64.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.64.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.65.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.65.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.65.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.66.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.66.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.66.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.67.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.67.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.67.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.68.down_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.68.gate_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.experts.68.up_proj.weight": "model-00078-of-00282.safetensors",
+    "model.layers.3.mlp.gate.e_score_correction_bias": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.69.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.69.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.69.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.7.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.7.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.7.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.70.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.70.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.70.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.71.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.71.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.71.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.72.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.72.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.72.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.73.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.73.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.73.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.74.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.74.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.74.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.75.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.75.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.75.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.76.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.76.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.76.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.77.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.77.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.77.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.78.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.78.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.78.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.79.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.79.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.79.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.8.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.8.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.8.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.80.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.80.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.80.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.81.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.81.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.81.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.82.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.82.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.82.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.83.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.83.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.83.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.84.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.84.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.84.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.85.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.85.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.85.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.86.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.86.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.86.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.87.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.87.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.87.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.88.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.88.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.88.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.89.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.89.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.89.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.9.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.9.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.9.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.90.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.90.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.90.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.91.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.91.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.91.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.92.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.92.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.92.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.93.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.93.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.93.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.94.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.94.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.94.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.95.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.95.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.95.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.96.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.96.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.96.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.97.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.97.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.97.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.98.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.98.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.98.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.99.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.99.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.experts.99.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.gate.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.shared_experts.down_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.shared_experts.gate_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.mlp.shared_experts.up_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.indexer.k_norm.bias": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.indexer.k_norm.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.indexer.weights_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.indexer.wk.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.indexer.wq_b.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.kv_a_layernorm.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.kv_a_proj_with_mqa.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.kv_b_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.q_a_layernorm.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.q_a_proj.weight": "model-00079-of-00282.safetensors",
+    "model.layers.3.self_attn.q_b_proj.weight": "model-00079-of-00282.safetensors",
+    "model.norm.weight": "model-00282-of-00282.safetensors"
+  }
+}

modeling_deepseek_v32.py

@@ -0,0 +1,339 @@
+import math
+import warnings
+from collections.abc import Callable
+from typing import Optional
+
+import torch
+import torch.nn.functional as F
+from torch import nn
+
+from transformers import initialization as init
+from transformers.cache_utils import Cache
+from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
+from transformers.modeling_layers import GenericForSequenceClassification, GenericForTokenClassification
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from transformers.processing_utils import Unpack
+from transformers.utils import logging
+from transformers.models.deepseek_v3.modeling_deepseek_v3 import (
+    DeepseekV3Attention,
+    DeepseekV3DecoderLayer,
+    DeepseekV3ForCausalLM,
+    DeepseekV3MLP,
+    DeepseekV3Model,
+    DeepseekV3MoE,
+    DeepseekV3PreTrainedModel,
+    DeepseekV3RMSNorm,
+    DeepseekV3RotaryEmbedding,
+    apply_rotary_pos_emb_interleave,
+    yarn_get_mscale,
+)
+from transformers.models.llama.modeling_llama import (
+    apply_rotary_pos_emb,
+    eager_attention_forward,
+)
+from configuration_deepseek_v32 import DeepseekV32Config
+
+
+logger = logging.get_logger(__name__)
+
+
+class DeepseekV32RMSNorm(DeepseekV3RMSNorm):
+    pass
+
+
+class DeepseekV32RotaryEmbedding(DeepseekV3RotaryEmbedding):
+    pass
+
+
+class DeepseekV32MLP(DeepseekV3MLP):
+    pass
+
+
+class DeepseekV32MoE(DeepseekV3MoE):
+    pass
+
+
+class DeepseekV32SparseAttention(nn.Module):
+    """
+    DeepSeek V3.2 sparse attention mechanism with indexer.
+    
+    This implements the native sparse attention from DeepSeek V3.2 which uses
+    an indexer to select top-k tokens for attention computation, making it
+    more efficient for long sequences.
+    """
+
+    def __init__(self, config: DeepseekV32Config, layer_idx: int):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.attention_dropout = config.attention_dropout
+        self.num_heads = config.num_attention_heads
+
+        self.q_lora_rank = config.q_lora_rank
+        self.qk_rope_head_dim = config.qk_rope_head_dim
+        self.kv_lora_rank = config.kv_lora_rank
+        self.v_head_dim = config.v_head_dim
+        self.qk_nope_head_dim = config.qk_nope_head_dim
+        self.qk_head_dim = config.qk_head_dim
+        self.index_topk = config.index_topk
+
+        self.is_causal = True
+
+        # Query projection
+        if self.q_lora_rank is None:
+            self.q_proj = nn.Linear(config.hidden_size, self.num_heads * self.qk_head_dim, bias=False)
+        else:
+            self.q_a_proj = nn.Linear(config.hidden_size, config.q_lora_rank, bias=config.attention_bias)
+            self.q_a_layernorm = DeepseekV32RMSNorm(config.q_lora_rank)
+            self.q_b_proj = nn.Linear(config.q_lora_rank, self.num_heads * self.qk_head_dim, bias=False)
+
+        # Key-Value projections
+        self.kv_a_proj_with_mqa = nn.Linear(
+            config.hidden_size,
+            self.kv_lora_rank + self.qk_rope_head_dim,
+            bias=config.attention_bias,
+        )
+        self.kv_a_layernorm = DeepseekV32RMSNorm(self.kv_lora_rank)
+        self.kv_b_proj = nn.Linear(
+            self.kv_lora_rank,
+            self.num_heads * (self.qk_nope_head_dim + self.v_head_dim),
+            bias=False,
+        )
+
+        # Output projection
+        self.o_proj = nn.Linear(
+            self.num_heads * self.v_head_dim,
+            config.hidden_size,
+            bias=config.attention_bias,
+        )
+
+        # Indexer components for sparse attention
+        self.wq_b = nn.Linear(config.q_lora_rank, self.num_heads * self.qk_head_dim, bias=False)
+        self.wk = nn.Linear(config.hidden_size, self.qk_head_dim, bias=config.attention_bias)
+        self.k_norm = DeepseekV32RMSNorm(self.qk_head_dim)
+        self.weights_proj = nn.Linear(config.hidden_size, self.num_heads, bias=False)
+
+        self.scaling = self.qk_head_dim ** (-0.5)
+        if self.config.rope_parameters.get("rope_type", "default") != "default":
+            mscale_all_dim = self.config.rope_parameters.get("mscale_all_dim", 0)
+            scaling_factor = self.config.rope_parameters["factor"]
+            if mscale_all_dim:
+                mscale = yarn_get_mscale(scaling_factor, mscale_all_dim)
+                self.scaling = self.scaling * mscale * mscale
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor],
+        past_key_values: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
+        batch_size, seq_length = hidden_states.shape[:-1]
+
+        # For training or when index_topk is not effective, fall back to standard attention
+        # This is a simplified implementation - in practice, you'd implement the full sparse indexer
+        if self.training or seq_length <= self.index_topk:
+            warnings.warn(
+                "DeepSeek V3.2 sparse attention is not fully implemented in this version. "
+                "Falling back to standard attention. For production use, please use vLLM or "
+                "other optimized inference engines.",
+                UserWarning,
+            )
+            return self._standard_attention(
+                hidden_states, position_embeddings, attention_mask, past_key_values, cache_position, **kwargs
+            )
+
+        # Sparse attention implementation would go here
+        # This requires custom CUDA kernels for efficient top-k selection and indexing
+        return self._standard_attention(
+            hidden_states, position_embeddings, attention_mask, past_key_values, cache_position, **kwargs
+        )
+
+    def _standard_attention(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor],
+        past_key_values: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
+        """Standard attention fallback (same as DeepSeek V3)"""
+        batch_size, seq_length = hidden_states.shape[:-1]
+        query_shape = (batch_size, seq_length, -1, self.qk_head_dim)
+        key_shape = (batch_size, seq_length, -1, self.qk_nope_head_dim + self.v_head_dim)
+
+        if self.q_lora_rank is None:
+            q_states = self.q_proj(hidden_states)
+        else:
+            q_states = self.q_b_proj(self.q_a_layernorm(self.q_a_proj(hidden_states)))
+        q_states = q_states.view(query_shape).transpose(1, 2)
+        q_pass, q_rot = torch.split(q_states, [self.qk_nope_head_dim, self.qk_rope_head_dim], dim=-1)
+
+        compressed_kv = self.kv_a_proj_with_mqa(hidden_states)
+        k_pass, k_rot = torch.split(compressed_kv, [self.kv_lora_rank, self.qk_rope_head_dim], dim=-1)
+
+        k_pass = self.kv_b_proj(self.kv_a_layernorm(k_pass)).view(key_shape).transpose(1, 2)
+        k_pass, value_states = torch.split(k_pass, [self.qk_nope_head_dim, self.v_head_dim], dim=-1)
+
+        k_rot = k_rot.view(batch_size, 1, seq_length, self.qk_rope_head_dim)
+
+        cos, sin = position_embeddings
+        if self.config.rope_interleave:
+            q_rot, k_rot = apply_rotary_pos_emb_interleave(q_rot, k_rot, cos, sin)
+        else:
+            q_rot, k_rot = apply_rotary_pos_emb(q_rot, k_rot, cos, sin)
+        k_rot = k_rot.expand(*k_pass.shape[:-1], -1)
+
+        query_states = torch.cat((q_pass, q_rot), dim=-1)
+        key_states = torch.cat((k_pass, k_rot), dim=-1)
+
+        if past_key_values is not None:
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+            key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim:
+            value_states = F.pad(value_states, [0, self.qk_head_dim - self.v_head_dim])
+
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            **kwargs,
+        )
+
+        if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim:
+            attn_output = attn_output[:, :, :, : self.v_head_dim]
+
+        attn_output = attn_output.reshape(batch_size, seq_length, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+class DeepseekV32DecoderLayer(nn.Module):
+    def __init__(self, config: DeepseekV32Config, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        # Use sparse attention for V3.2
+        self.self_attn = DeepseekV32SparseAttention(config=config, layer_idx=layer_idx)
+
+        if layer_idx >= config.first_k_dense_replace:
+            self.mlp = DeepseekV32MoE(config)
+        else:
+            self.mlp = DeepseekV32MLP(config)
+
+        self.input_layernorm = DeepseekV32RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = DeepseekV32RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        past_key_values: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> torch.Tensor:
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            position_embeddings=position_embeddings,
+            attention_mask=attention_mask,
+            past_key_values=past_key_values,
+            cache_position=cache_position,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        return hidden_states
+
+
+class DeepseekV32PreTrainedModel(DeepseekV3PreTrainedModel):
+    config_class = DeepseekV32Config
+    _can_compile_fullgraph = False
+    _keep_in_fp32_modules_strict = ["e_score_correction_bias"]
+
+
+class DeepseekV32Model(DeepseekV3Model):
+    """
+    DeepSeek V3.2 Model with native sparse attention.
+    
+    This model extends DeepSeek V3 with an efficient sparse attention mechanism
+    that uses an indexer to select top-k tokens for attention computation.
+    """
+    config_class = DeepseekV32Config
+    _keys_to_ignore_on_load_unexpected = [r"model\.layers\.61.*"]
+
+    def __init__(self, config: DeepseekV32Config):
+        # Skip DeepseekV3Model.__init__ and go directly to PreTrainedModel
+        DeepseekV3PreTrainedModel.__init__(self, config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        # Use V3.2-specific decoder layers
+        self.layers = nn.ModuleList(
+            [DeepseekV32DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = DeepseekV32RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = DeepseekV32RotaryEmbedding(config=config)
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+
+class DeepseekV32ForCausalLM(DeepseekV3ForCausalLM):
+    """
+    DeepSeek V3.2 Model for causal language modeling with sparse attention.
+    """
+    config_class = DeepseekV32Config
+    _tied_weights_keys = ["lm_head.weight"]
+
+    def __init__(self, config):
+        super(DeepseekV3ForCausalLM, self).__init__(config)
+        self.model = DeepseekV32Model(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+
+class DeepseekV32ForSequenceClassification(GenericForSequenceClassification, DeepseekV32PreTrainedModel):
+    pass
+
+
+class DeepseekV32ForTokenClassification(GenericForTokenClassification, DeepseekV32PreTrainedModel):
+    pass
+
+
+__all__ = [
+    "DeepseekV32PreTrainedModel",
+    "DeepseekV32Model",
+    "DeepseekV32ForCausalLM",
+    "DeepseekV32ForSequenceClassification",
+    "DeepseekV32ForTokenClassification",
+]

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19e773648cb4e65de8660ea6365e10acca112d42a854923df93db4a6f333a82d
+size 20217442

tokenizer_config.json

@@ -0,0 +1,33 @@
+{
+  "backend": "tokenizers",
+  "clean_up_tokenization_spaces": false,
+  "do_lower_case": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": [
+    "<|endoftext|>",
+    "[MASK]",
+    "[gMASK]",
+    "[sMASK]",
+    "<sop>",
+    "<eop>",
+    "<|system|>",
+    "<|user|>",
+    "<|assistant|>",
+    "<|observation|>",
+    "<|begin_of_image|>",
+    "<|end_of_image|>",
+    "<|begin_of_video|>",
+    "<|end_of_video|>",
+    "<|begin_of_audio|>",
+    "<|end_of_audio|>",
+    "<|begin_of_transcription|>",
+    "<|end_of_transcription|>"
+  ],
+  "is_local": true,
+  "model_max_length": 202752,
+  "model_specific_special_tokens": {},
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "remove_space": false,
+  "tokenizer_class": "TokenizersBackend"
+}