Duplicate from skyblanket/glm5-abliterated-fp8

Co-authored-by: skyblanket <skyblanket@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,37 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz 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
+model.safetensors.index.json filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,45 @@
+---
+license: other
+base_model: THUDM/GLM-5-FP8
+tags:
+  - abliteration
+  - uncensored
+  - glm5
+  - moe
+  - fp8
+model_type: glm-moe
+---
+
+# GLM-5 744B Abliterated (FP8)
+
+This doesnt work an abliterated (uncensored) version of [THUDM/GLM-5-FP8](https://huggingface.co/THUDM/GLM-5-FP8) with safety guardrails removed via weight orthogonalization.
+
+## Method
+
+**Abliteration** (representation engineering) was used to identify and remove the "refusal direction" from the model's residual stream:
+
+1. **Computed refusal directions** for all 78 layers by collecting activations on 50 harmful vs 50 harmless prompts and computing mean difference vectors
+2. **Applied weight orthogonalization** to layers 15-54 (o_proj and shared_experts.down_proj) with alpha=1.0
+3. **FP8-aware processing**: Proper dequantization using block-wise scale_inv factors, abliteration in float32, and re-quantization preserving original scale factors to minimize perturbation
+
+### Technical Details
+
+- **Architecture**: GLM-5 MoE (744B total, 40B active), 78 layers, 6144 hidden dim
+- **Layers 0-2**: Dense MLP, **Layers 3-77**: MoE with FP8Expert fused kernels
+- **Modified weights**: 80 weight matrices (40 o_proj + 40 shared_experts.down_proj)
+- **Quantization**: FP8 E4M3 with block-wise scaling (128x128 blocks)
+- **Scale preservation**: Original weight_scale_inv factors retained for minimal quantization drift
+
+### Hardware Used
+
+8x NVIDIA B200 (1.4TB VRAM) on Vast.ai
+
+## Usage
+
+This model requires the same setup as the base GLM-5-FP8 model. Use `trust_remote_code=True` when loading.
+
+
+
+## Disclaimer
+
+This model is provided for research purposes only. The removal of safety guardrails means the model may generate harmful, biased, or offensive content. Users are responsible for ensuring appropriate use.

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 +1,863 @@
+{
+  "architectures": [
+    "GlmMoeDsaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 0,
+  "dtype": "bfloat16",
+  "eos_token_id": [
+    154820,
+    154827,
+    154829
+  ],
+  "ep_size": 1,
+  "first_k_dense_replace": 3,
+  "head_dim": 64,
+  "hidden_act": "silu",
+  "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,
+  "mlp_layer_types": [
+    "dense",
+    "dense",
+    "dense",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse"
+  ],
+  "model_type": "glm_moe_dsa",
+  "moe_intermediate_size": 2048,
+  "moe_layer_freq": 1,
+  "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": 78,
+  "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,
+  "quantization_config": {
+    "activation_scheme": "dynamic",
+    "dequantize": false,
+    "modules_to_not_convert": [
+      "lm_head",
+      "model.embed_tokens",
+      "model.layers.0.input_layernorm",
+      "model.layers.0.post_attention_layernorm",
+      "model.layers.0.self_attn.indexer.k_norm",
+      "model.layers.0.self_attn.indexer.k_norm.bias",
+      "model.layers.0.self_attn.indexers_proj",
+      "model.layers.0.self_attn.kv_a_layernorm",
+      "model.layers.0.self_attn.q_a_layernorm",
+      "model.layers.1.input_layernorm",
+      "model.layers.1.post_attention_layernorm",
+      "model.layers.1.self_attn.indexer.k_norm",
+      "model.layers.1.self_attn.indexer.k_norm.bias",
+      "model.layers.1.self_attn.indexers_proj",
+      "model.layers.1.self_attn.kv_a_layernorm",
+      "model.layers.1.self_attn.q_a_layernorm",
+      "model.layers.2.input_layernorm",
+      "model.layers.2.post_attention_layernorm",
+      "model.layers.2.self_attn.indexer.k_norm",
+      "model.layers.2.self_attn.indexer.k_norm.bias",
+      "model.layers.2.self_attn.indexers_proj",
+      "model.layers.2.self_attn.kv_a_layernorm",
+      "model.layers.2.self_attn.q_a_layernorm",
+      "model.layers.3.input_layernorm",
+      "model.layers.3.mlp.gate",
+      "model.layers.3.mlp.gate.e_score_correction_bias",
+      "model.layers.3.post_attention_layernorm",
+      "model.layers.3.self_attn.indexer.k_norm",
+      "model.layers.3.self_attn.indexer.k_norm.bias",
+      "model.layers.3.self_attn.indexers_proj",
+      "model.layers.3.self_attn.kv_a_layernorm",
+      "model.layers.3.self_attn.q_a_layernorm",
+      "model.layers.4.input_layernorm",
+      "model.layers.4.mlp.gate",
+      "model.layers.4.mlp.gate.e_score_correction_bias",
+      "model.layers.4.post_attention_layernorm",
+      "model.layers.4.self_attn.indexer.k_norm",
+      "model.layers.4.self_attn.indexer.k_norm.bias",
+      "model.layers.4.self_attn.indexers_proj",
+      "model.layers.4.self_attn.kv_a_layernorm",
+      "model.layers.4.self_attn.q_a_layernorm",
+      "model.layers.5.input_layernorm",
+      "model.layers.5.mlp.gate",
+      "model.layers.5.mlp.gate.e_score_correction_bias",
+      "model.layers.5.post_attention_layernorm",
+      "model.layers.5.self_attn.indexer.k_norm",
+      "model.layers.5.self_attn.indexer.k_norm.bias",
+      "model.layers.5.self_attn.indexers_proj",
+      "model.layers.5.self_attn.kv_a_layernorm",
+      "model.layers.5.self_attn.q_a_layernorm",
+      "model.layers.6.input_layernorm",
+      "model.layers.6.mlp.gate",
+      "model.layers.6.mlp.gate.e_score_correction_bias",
+      "model.layers.6.post_attention_layernorm",
+      "model.layers.6.self_attn.indexer.k_norm",
+      "model.layers.6.self_attn.indexer.k_norm.bias",
+      "model.layers.6.self_attn.indexers_proj",
+      "model.layers.6.self_attn.kv_a_layernorm",
+      "model.layers.6.self_attn.q_a_layernorm",
+      "model.layers.7.input_layernorm",
+      "model.layers.7.mlp.gate",
+      "model.layers.7.mlp.gate.e_score_correction_bias",
+      "model.layers.7.post_attention_layernorm",
+      "model.layers.7.self_attn.indexer.k_norm",
+      "model.layers.7.self_attn.indexer.k_norm.bias",
+      "model.layers.7.self_attn.indexers_proj",
+      "model.layers.7.self_attn.kv_a_layernorm",
+      "model.layers.7.self_attn.q_a_layernorm",
+      "model.layers.8.input_layernorm",
+      "model.layers.8.mlp.gate",
+      "model.layers.8.mlp.gate.e_score_correction_bias",
+      "model.layers.8.post_attention_layernorm",
+      "model.layers.8.self_attn.indexer.k_norm",
+      "model.layers.8.self_attn.indexer.k_norm.bias",
+      "model.layers.8.self_attn.indexers_proj",
+      "model.layers.8.self_attn.kv_a_layernorm",
+      "model.layers.8.self_attn.q_a_layernorm",
+      "model.layers.9.input_layernorm",
+      "model.layers.9.mlp.gate",
+      "model.layers.9.mlp.gate.e_score_correction_bias",
+      "model.layers.9.post_attention_layernorm",
+      "model.layers.9.self_attn.indexer.k_norm",
+      "model.layers.9.self_attn.indexer.k_norm.bias",
+      "model.layers.9.self_attn.indexers_proj",
+      "model.layers.9.self_attn.kv_a_layernorm",
+      "model.layers.9.self_attn.q_a_layernorm",
+      "model.layers.10.input_layernorm",
+      "model.layers.10.mlp.gate",
+      "model.layers.10.mlp.gate.e_score_correction_bias",
+      "model.layers.10.post_attention_layernorm",
+      "model.layers.10.self_attn.indexer.k_norm",
+      "model.layers.10.self_attn.indexer.k_norm.bias",
+      "model.layers.10.self_attn.indexers_proj",
+      "model.layers.10.self_attn.kv_a_layernorm",
+      "model.layers.10.self_attn.q_a_layernorm",
+      "model.layers.11.input_layernorm",
+      "model.layers.11.mlp.gate",
+      "model.layers.11.mlp.gate.e_score_correction_bias",
+      "model.layers.11.post_attention_layernorm",
+      "model.layers.11.self_attn.indexer.k_norm",
+      "model.layers.11.self_attn.indexer.k_norm.bias",
+      "model.layers.11.self_attn.indexers_proj",
+      "model.layers.11.self_attn.kv_a_layernorm",
+      "model.layers.11.self_attn.q_a_layernorm",
+      "model.layers.12.input_layernorm",
+      "model.layers.12.mlp.gate",
+      "model.layers.12.mlp.gate.e_score_correction_bias",
+      "model.layers.12.post_attention_layernorm",
+      "model.layers.12.self_attn.indexer.k_norm",
+      "model.layers.12.self_attn.indexer.k_norm.bias",
+      "model.layers.12.self_attn.indexers_proj",
+      "model.layers.12.self_attn.kv_a_layernorm",
+      "model.layers.12.self_attn.q_a_layernorm",
+      "model.layers.13.input_layernorm",
+      "model.layers.13.mlp.gate",
+      "model.layers.13.mlp.gate.e_score_correction_bias",
+      "model.layers.13.post_attention_layernorm",
+      "model.layers.13.self_attn.indexer.k_norm",
+      "model.layers.13.self_attn.indexer.k_norm.bias",
+      "model.layers.13.self_attn.indexers_proj",
+      "model.layers.13.self_attn.kv_a_layernorm",
+      "model.layers.13.self_attn.q_a_layernorm",
+      "model.layers.14.input_layernorm",
+      "model.layers.14.mlp.gate",
+      "model.layers.14.mlp.gate.e_score_correction_bias",
+      "model.layers.14.post_attention_layernorm",
+      "model.layers.14.self_attn.indexer.k_norm",
+      "model.layers.14.self_attn.indexer.k_norm.bias",
+      "model.layers.14.self_attn.indexers_proj",
+      "model.layers.14.self_attn.kv_a_layernorm",
+      "model.layers.14.self_attn.q_a_layernorm",
+      "model.layers.15.input_layernorm",
+      "model.layers.15.mlp.gate",
+      "model.layers.15.mlp.gate.e_score_correction_bias",
+      "model.layers.15.post_attention_layernorm",
+      "model.layers.15.self_attn.indexer.k_norm",
+      "model.layers.15.self_attn.indexer.k_norm.bias",
+      "model.layers.15.self_attn.indexers_proj",
+      "model.layers.15.self_attn.kv_a_layernorm",
+      "model.layers.15.self_attn.q_a_layernorm",
+      "model.layers.16.input_layernorm",
+      "model.layers.16.mlp.gate",
+      "model.layers.16.mlp.gate.e_score_correction_bias",
+      "model.layers.16.post_attention_layernorm",
+      "model.layers.16.self_attn.indexer.k_norm",
+      "model.layers.16.self_attn.indexer.k_norm.bias",
+      "model.layers.16.self_attn.indexers_proj",
+      "model.layers.16.self_attn.kv_a_layernorm",
+      "model.layers.16.self_attn.q_a_layernorm",
+      "model.layers.17.input_layernorm",
+      "model.layers.17.mlp.gate",
+      "model.layers.17.mlp.gate.e_score_correction_bias",
+      "model.layers.17.post_attention_layernorm",
+      "model.layers.17.self_attn.indexer.k_norm",
+      "model.layers.17.self_attn.indexer.k_norm.bias",
+      "model.layers.17.self_attn.indexers_proj",
+      "model.layers.17.self_attn.kv_a_layernorm",
+      "model.layers.17.self_attn.q_a_layernorm",
+      "model.layers.18.input_layernorm",
+      "model.layers.18.mlp.gate",
+      "model.layers.18.mlp.gate.e_score_correction_bias",
+      "model.layers.18.post_attention_layernorm",
+      "model.layers.18.self_attn.indexer.k_norm",
+      "model.layers.18.self_attn.indexer.k_norm.bias",
+      "model.layers.18.self_attn.indexers_proj",
+      "model.layers.18.self_attn.kv_a_layernorm",
+      "model.layers.18.self_attn.q_a_layernorm",
+      "model.layers.19.input_layernorm",
+      "model.layers.19.mlp.gate",
+      "model.layers.19.mlp.gate.e_score_correction_bias",
+      "model.layers.19.post_attention_layernorm",
+      "model.layers.19.self_attn.indexer.k_norm",
+      "model.layers.19.self_attn.indexer.k_norm.bias",
+      "model.layers.19.self_attn.indexers_proj",
+      "model.layers.19.self_attn.kv_a_layernorm",
+      "model.layers.19.self_attn.q_a_layernorm",
+      "model.layers.20.input_layernorm",
+      "model.layers.20.mlp.gate",
+      "model.layers.20.mlp.gate.e_score_correction_bias",
+      "model.layers.20.post_attention_layernorm",
+      "model.layers.20.self_attn.indexer.k_norm",
+      "model.layers.20.self_attn.indexer.k_norm.bias",
+      "model.layers.20.self_attn.indexers_proj",
+      "model.layers.20.self_attn.kv_a_layernorm",
+      "model.layers.20.self_attn.q_a_layernorm",
+      "model.layers.21.input_layernorm",
+      "model.layers.21.mlp.gate",
+      "model.layers.21.mlp.gate.e_score_correction_bias",
+      "model.layers.21.post_attention_layernorm",
+      "model.layers.21.self_attn.indexer.k_norm",
+      "model.layers.21.self_attn.indexer.k_norm.bias",
+      "model.layers.21.self_attn.indexers_proj",
+      "model.layers.21.self_attn.kv_a_layernorm",
+      "model.layers.21.self_attn.q_a_layernorm",
+      "model.layers.22.input_layernorm",
+      "model.layers.22.mlp.gate",
+      "model.layers.22.mlp.gate.e_score_correction_bias",
+      "model.layers.22.post_attention_layernorm",
+      "model.layers.22.self_attn.indexer.k_norm",
+      "model.layers.22.self_attn.indexer.k_norm.bias",
+      "model.layers.22.self_attn.indexers_proj",
+      "model.layers.22.self_attn.kv_a_layernorm",
+      "model.layers.22.self_attn.q_a_layernorm",
+      "model.layers.23.input_layernorm",
+      "model.layers.23.mlp.gate",
+      "model.layers.23.mlp.gate.e_score_correction_bias",
+      "model.layers.23.post_attention_layernorm",
+      "model.layers.23.self_attn.indexer.k_norm",
+      "model.layers.23.self_attn.indexer.k_norm.bias",
+      "model.layers.23.self_attn.indexers_proj",
+      "model.layers.23.self_attn.kv_a_layernorm",
+      "model.layers.23.self_attn.q_a_layernorm",
+      "model.layers.24.input_layernorm",
+      "model.layers.24.mlp.gate",
+      "model.layers.24.mlp.gate.e_score_correction_bias",
+      "model.layers.24.post_attention_layernorm",
+      "model.layers.24.self_attn.indexer.k_norm",
+      "model.layers.24.self_attn.indexer.k_norm.bias",
+      "model.layers.24.self_attn.indexers_proj",
+      "model.layers.24.self_attn.kv_a_layernorm",
+      "model.layers.24.self_attn.q_a_layernorm",
+      "model.layers.25.input_layernorm",
+      "model.layers.25.mlp.gate",
+      "model.layers.25.mlp.gate.e_score_correction_bias",
+      "model.layers.25.post_attention_layernorm",
+      "model.layers.25.self_attn.indexer.k_norm",
+      "model.layers.25.self_attn.indexer.k_norm.bias",
+      "model.layers.25.self_attn.indexers_proj",
+      "model.layers.25.self_attn.kv_a_layernorm",
+      "model.layers.25.self_attn.q_a_layernorm",
+      "model.layers.26.input_layernorm",
+      "model.layers.26.mlp.gate",
+      "model.layers.26.mlp.gate.e_score_correction_bias",
+      "model.layers.26.post_attention_layernorm",
+      "model.layers.26.self_attn.indexer.k_norm",
+      "model.layers.26.self_attn.indexer.k_norm.bias",
+      "model.layers.26.self_attn.indexers_proj",
+      "model.layers.26.self_attn.kv_a_layernorm",
+      "model.layers.26.self_attn.q_a_layernorm",
+      "model.layers.27.input_layernorm",
+      "model.layers.27.mlp.gate",
+      "model.layers.27.mlp.gate.e_score_correction_bias",
+      "model.layers.27.post_attention_layernorm",
+      "model.layers.27.self_attn.indexer.k_norm",
+      "model.layers.27.self_attn.indexer.k_norm.bias",
+      "model.layers.27.self_attn.indexers_proj",
+      "model.layers.27.self_attn.kv_a_layernorm",
+      "model.layers.27.self_attn.q_a_layernorm",
+      "model.layers.28.input_layernorm",
+      "model.layers.28.mlp.gate",
+      "model.layers.28.mlp.gate.e_score_correction_bias",
+      "model.layers.28.post_attention_layernorm",
+      "model.layers.28.self_attn.indexer.k_norm",
+      "model.layers.28.self_attn.indexer.k_norm.bias",
+      "model.layers.28.self_attn.indexers_proj",
+      "model.layers.28.self_attn.kv_a_layernorm",
+      "model.layers.28.self_attn.q_a_layernorm",
+      "model.layers.29.input_layernorm",
+      "model.layers.29.mlp.gate",
+      "model.layers.29.mlp.gate.e_score_correction_bias",
+      "model.layers.29.post_attention_layernorm",
+      "model.layers.29.self_attn.indexer.k_norm",
+      "model.layers.29.self_attn.indexer.k_norm.bias",
+      "model.layers.29.self_attn.indexers_proj",
+      "model.layers.29.self_attn.kv_a_layernorm",
+      "model.layers.29.self_attn.q_a_layernorm",
+      "model.layers.30.input_layernorm",
+      "model.layers.30.mlp.gate",
+      "model.layers.30.mlp.gate.e_score_correction_bias",
+      "model.layers.30.post_attention_layernorm",
+      "model.layers.30.self_attn.indexer.k_norm",
+      "model.layers.30.self_attn.indexer.k_norm.bias",
+      "model.layers.30.self_attn.indexers_proj",
+      "model.layers.30.self_attn.kv_a_layernorm",
+      "model.layers.30.self_attn.q_a_layernorm",
+      "model.layers.31.input_layernorm",
+      "model.layers.31.mlp.gate",
+      "model.layers.31.mlp.gate.e_score_correction_bias",
+      "model.layers.31.post_attention_layernorm",
+      "model.layers.31.self_attn.indexer.k_norm",
+      "model.layers.31.self_attn.indexer.k_norm.bias",
+      "model.layers.31.self_attn.indexers_proj",
+      "model.layers.31.self_attn.kv_a_layernorm",
+      "model.layers.31.self_attn.q_a_layernorm",
+      "model.layers.32.input_layernorm",
+      "model.layers.32.mlp.gate",
+      "model.layers.32.mlp.gate.e_score_correction_bias",
+      "model.layers.32.post_attention_layernorm",
+      "model.layers.32.self_attn.indexer.k_norm",
+      "model.layers.32.self_attn.indexer.k_norm.bias",
+      "model.layers.32.self_attn.indexers_proj",
+      "model.layers.32.self_attn.kv_a_layernorm",
+      "model.layers.32.self_attn.q_a_layernorm",
+      "model.layers.33.input_layernorm",
+      "model.layers.33.mlp.gate",
+      "model.layers.33.mlp.gate.e_score_correction_bias",
+      "model.layers.33.post_attention_layernorm",
+      "model.layers.33.self_attn.indexer.k_norm",
+      "model.layers.33.self_attn.indexer.k_norm.bias",
+      "model.layers.33.self_attn.indexers_proj",
+      "model.layers.33.self_attn.kv_a_layernorm",
+      "model.layers.33.self_attn.q_a_layernorm",
+      "model.layers.34.input_layernorm",
+      "model.layers.34.mlp.gate",
+      "model.layers.34.mlp.gate.e_score_correction_bias",
+      "model.layers.34.post_attention_layernorm",
+      "model.layers.34.self_attn.indexer.k_norm",
+      "model.layers.34.self_attn.indexer.k_norm.bias",
+      "model.layers.34.self_attn.indexers_proj",
+      "model.layers.34.self_attn.kv_a_layernorm",
+      "model.layers.34.self_attn.q_a_layernorm",
+      "model.layers.35.input_layernorm",
+      "model.layers.35.mlp.gate",
+      "model.layers.35.mlp.gate.e_score_correction_bias",
+      "model.layers.35.post_attention_layernorm",
+      "model.layers.35.self_attn.indexer.k_norm",
+      "model.layers.35.self_attn.indexer.k_norm.bias",
+      "model.layers.35.self_attn.indexers_proj",
+      "model.layers.35.self_attn.kv_a_layernorm",
+      "model.layers.35.self_attn.q_a_layernorm",
+      "model.layers.36.input_layernorm",
+      "model.layers.36.mlp.gate",
+      "model.layers.36.mlp.gate.e_score_correction_bias",
+      "model.layers.36.post_attention_layernorm",
+      "model.layers.36.self_attn.indexer.k_norm",
+      "model.layers.36.self_attn.indexer.k_norm.bias",
+      "model.layers.36.self_attn.indexers_proj",
+      "model.layers.36.self_attn.kv_a_layernorm",
+      "model.layers.36.self_attn.q_a_layernorm",
+      "model.layers.37.input_layernorm",
+      "model.layers.37.mlp.gate",
+      "model.layers.37.mlp.gate.e_score_correction_bias",
+      "model.layers.37.post_attention_layernorm",
+      "model.layers.37.self_attn.indexer.k_norm",
+      "model.layers.37.self_attn.indexer.k_norm.bias",
+      "model.layers.37.self_attn.indexers_proj",
+      "model.layers.37.self_attn.kv_a_layernorm",
+      "model.layers.37.self_attn.q_a_layernorm",
+      "model.layers.38.input_layernorm",
+      "model.layers.38.mlp.gate",
+      "model.layers.38.mlp.gate.e_score_correction_bias",
+      "model.layers.38.post_attention_layernorm",
+      "model.layers.38.self_attn.indexer.k_norm",
+      "model.layers.38.self_attn.indexer.k_norm.bias",
+      "model.layers.38.self_attn.indexers_proj",
+      "model.layers.38.self_attn.kv_a_layernorm",
+      "model.layers.38.self_attn.q_a_layernorm",
+      "model.layers.39.input_layernorm",
+      "model.layers.39.mlp.gate",
+      "model.layers.39.mlp.gate.e_score_correction_bias",
+      "model.layers.39.post_attention_layernorm",
+      "model.layers.39.self_attn.indexer.k_norm",
+      "model.layers.39.self_attn.indexer.k_norm.bias",
+      "model.layers.39.self_attn.indexers_proj",
+      "model.layers.39.self_attn.kv_a_layernorm",
+      "model.layers.39.self_attn.q_a_layernorm",
+      "model.layers.40.input_layernorm",
+      "model.layers.40.mlp.gate",
+      "model.layers.40.mlp.gate.e_score_correction_bias",
+      "model.layers.40.post_attention_layernorm",
+      "model.layers.40.self_attn.indexer.k_norm",
+      "model.layers.40.self_attn.indexer.k_norm.bias",
+      "model.layers.40.self_attn.indexers_proj",
+      "model.layers.40.self_attn.kv_a_layernorm",
+      "model.layers.40.self_attn.q_a_layernorm",
+      "model.layers.41.input_layernorm",
+      "model.layers.41.mlp.gate",
+      "model.layers.41.mlp.gate.e_score_correction_bias",
+      "model.layers.41.post_attention_layernorm",
+      "model.layers.41.self_attn.indexer.k_norm",
+      "model.layers.41.self_attn.indexer.k_norm.bias",
+      "model.layers.41.self_attn.indexers_proj",
+      "model.layers.41.self_attn.kv_a_layernorm",
+      "model.layers.41.self_attn.q_a_layernorm",
+      "model.layers.42.input_layernorm",
+      "model.layers.42.mlp.gate",
+      "model.layers.42.mlp.gate.e_score_correction_bias",
+      "model.layers.42.post_attention_layernorm",
+      "model.layers.42.self_attn.indexer.k_norm",
+      "model.layers.42.self_attn.indexer.k_norm.bias",
+      "model.layers.42.self_attn.indexers_proj",
+      "model.layers.42.self_attn.kv_a_layernorm",
+      "model.layers.42.self_attn.q_a_layernorm",
+      "model.layers.43.input_layernorm",
+      "model.layers.43.mlp.gate",
+      "model.layers.43.mlp.gate.e_score_correction_bias",
+      "model.layers.43.post_attention_layernorm",
+      "model.layers.43.self_attn.indexer.k_norm",
+      "model.layers.43.self_attn.indexer.k_norm.bias",
+      "model.layers.43.self_attn.indexers_proj",
+      "model.layers.43.self_attn.kv_a_layernorm",
+      "model.layers.43.self_attn.q_a_layernorm",
+      "model.layers.44.input_layernorm",
+      "model.layers.44.mlp.gate",
+      "model.layers.44.mlp.gate.e_score_correction_bias",
+      "model.layers.44.post_attention_layernorm",
+      "model.layers.44.self_attn.indexer.k_norm",
+      "model.layers.44.self_attn.indexer.k_norm.bias",
+      "model.layers.44.self_attn.indexers_proj",
+      "model.layers.44.self_attn.kv_a_layernorm",
+      "model.layers.44.self_attn.q_a_layernorm",
+      "model.layers.45.input_layernorm",
+      "model.layers.45.mlp.gate",
+      "model.layers.45.mlp.gate.e_score_correction_bias",
+      "model.layers.45.post_attention_layernorm",
+      "model.layers.45.self_attn.indexer.k_norm",
+      "model.layers.45.self_attn.indexer.k_norm.bias",
+      "model.layers.45.self_attn.indexers_proj",
+      "model.layers.45.self_attn.kv_a_layernorm",
+      "model.layers.45.self_attn.q_a_layernorm",
+      "model.layers.46.input_layernorm",
+      "model.layers.46.mlp.gate",
+      "model.layers.46.mlp.gate.e_score_correction_bias",
+      "model.layers.46.post_attention_layernorm",
+      "model.layers.46.self_attn.indexer.k_norm",
+      "model.layers.46.self_attn.indexer.k_norm.bias",
+      "model.layers.46.self_attn.indexers_proj",
+      "model.layers.46.self_attn.kv_a_layernorm",
+      "model.layers.46.self_attn.q_a_layernorm",
+      "model.layers.47.input_layernorm",
+      "model.layers.47.mlp.gate",
+      "model.layers.47.mlp.gate.e_score_correction_bias",
+      "model.layers.47.post_attention_layernorm",
+      "model.layers.47.self_attn.indexer.k_norm",
+      "model.layers.47.self_attn.indexer.k_norm.bias",
+      "model.layers.47.self_attn.indexers_proj",
+      "model.layers.47.self_attn.kv_a_layernorm",
+      "model.layers.47.self_attn.q_a_layernorm",
+      "model.layers.48.input_layernorm",
+      "model.layers.48.mlp.gate",
+      "model.layers.48.mlp.gate.e_score_correction_bias",
+      "model.layers.48.post_attention_layernorm",
+      "model.layers.48.self_attn.indexer.k_norm",
+      "model.layers.48.self_attn.indexer.k_norm.bias",
+      "model.layers.48.self_attn.indexers_proj",
+      "model.layers.48.self_attn.kv_a_layernorm",
+      "model.layers.48.self_attn.q_a_layernorm",
+      "model.layers.49.input_layernorm",
+      "model.layers.49.mlp.gate",
+      "model.layers.49.mlp.gate.e_score_correction_bias",
+      "model.layers.49.post_attention_layernorm",
+      "model.layers.49.self_attn.indexer.k_norm",
+      "model.layers.49.self_attn.indexer.k_norm.bias",
+      "model.layers.49.self_attn.indexers_proj",
+      "model.layers.49.self_attn.kv_a_layernorm",
+      "model.layers.49.self_attn.q_a_layernorm",
+      "model.layers.50.input_layernorm",
+      "model.layers.50.mlp.gate",
+      "model.layers.50.mlp.gate.e_score_correction_bias",
+      "model.layers.50.post_attention_layernorm",
+      "model.layers.50.self_attn.indexer.k_norm",
+      "model.layers.50.self_attn.indexer.k_norm.bias",
+      "model.layers.50.self_attn.indexers_proj",
+      "model.layers.50.self_attn.kv_a_layernorm",
+      "model.layers.50.self_attn.q_a_layernorm",
+      "model.layers.51.input_layernorm",
+      "model.layers.51.mlp.gate",
+      "model.layers.51.mlp.gate.e_score_correction_bias",
+      "model.layers.51.post_attention_layernorm",
+      "model.layers.51.self_attn.indexer.k_norm",
+      "model.layers.51.self_attn.indexer.k_norm.bias",
+      "model.layers.51.self_attn.indexers_proj",
+      "model.layers.51.self_attn.kv_a_layernorm",
+      "model.layers.51.self_attn.q_a_layernorm",
+      "model.layers.52.input_layernorm",
+      "model.layers.52.mlp.gate",
+      "model.layers.52.mlp.gate.e_score_correction_bias",
+      "model.layers.52.post_attention_layernorm",
+      "model.layers.52.self_attn.indexer.k_norm",
+      "model.layers.52.self_attn.indexer.k_norm.bias",
+      "model.layers.52.self_attn.indexers_proj",
+      "model.layers.52.self_attn.kv_a_layernorm",
+      "model.layers.52.self_attn.q_a_layernorm",
+      "model.layers.53.input_layernorm",
+      "model.layers.53.mlp.gate",
+      "model.layers.53.mlp.gate.e_score_correction_bias",
+      "model.layers.53.post_attention_layernorm",
+      "model.layers.53.self_attn.indexer.k_norm",
+      "model.layers.53.self_attn.indexer.k_norm.bias",
+      "model.layers.53.self_attn.indexers_proj",
+      "model.layers.53.self_attn.kv_a_layernorm",
+      "model.layers.53.self_attn.q_a_layernorm",
+      "model.layers.54.input_layernorm",
+      "model.layers.54.mlp.gate",
+      "model.layers.54.mlp.gate.e_score_correction_bias",
+      "model.layers.54.post_attention_layernorm",
+      "model.layers.54.self_attn.indexer.k_norm",
+      "model.layers.54.self_attn.indexer.k_norm.bias",
+      "model.layers.54.self_attn.indexers_proj",
+      "model.layers.54.self_attn.kv_a_layernorm",
+      "model.layers.54.self_attn.q_a_layernorm",
+      "model.layers.55.input_layernorm",
+      "model.layers.55.mlp.gate",
+      "model.layers.55.mlp.gate.e_score_correction_bias",
+      "model.layers.55.post_attention_layernorm",
+      "model.layers.55.self_attn.indexer.k_norm",
+      "model.layers.55.self_attn.indexer.k_norm.bias",
+      "model.layers.55.self_attn.indexers_proj",
+      "model.layers.55.self_attn.kv_a_layernorm",
+      "model.layers.55.self_attn.q_a_layernorm",
+      "model.layers.56.input_layernorm",
+      "model.layers.56.mlp.gate",
+      "model.layers.56.mlp.gate.e_score_correction_bias",
+      "model.layers.56.post_attention_layernorm",
+      "model.layers.56.self_attn.indexer.k_norm",
+      "model.layers.56.self_attn.indexer.k_norm.bias",
+      "model.layers.56.self_attn.indexers_proj",
+      "model.layers.56.self_attn.kv_a_layernorm",
+      "model.layers.56.self_attn.q_a_layernorm",
+      "model.layers.57.input_layernorm",
+      "model.layers.57.mlp.gate",
+      "model.layers.57.mlp.gate.e_score_correction_bias",
+      "model.layers.57.post_attention_layernorm",
+      "model.layers.57.self_attn.indexer.k_norm",
+      "model.layers.57.self_attn.indexer.k_norm.bias",
+      "model.layers.57.self_attn.indexers_proj",
+      "model.layers.57.self_attn.kv_a_layernorm",
+      "model.layers.57.self_attn.q_a_layernorm",
+      "model.layers.58.input_layernorm",
+      "model.layers.58.mlp.gate",
+      "model.layers.58.mlp.gate.e_score_correction_bias",
+      "model.layers.58.post_attention_layernorm",
+      "model.layers.58.self_attn.indexer.k_norm",
+      "model.layers.58.self_attn.indexer.k_norm.bias",
+      "model.layers.58.self_attn.indexers_proj",
+      "model.layers.58.self_attn.kv_a_layernorm",
+      "model.layers.58.self_attn.q_a_layernorm",
+      "model.layers.59.input_layernorm",
+      "model.layers.59.mlp.gate",
+      "model.layers.59.mlp.gate.e_score_correction_bias",
+      "model.layers.59.post_attention_layernorm",
+      "model.layers.59.self_attn.indexer.k_norm",
+      "model.layers.59.self_attn.indexer.k_norm.bias",
+      "model.layers.59.self_attn.indexers_proj",
+      "model.layers.59.self_attn.kv_a_layernorm",
+      "model.layers.59.self_attn.q_a_layernorm",
+      "model.layers.60.input_layernorm",
+      "model.layers.60.mlp.gate",
+      "model.layers.60.mlp.gate.e_score_correction_bias",
+      "model.layers.60.post_attention_layernorm",
+      "model.layers.60.self_attn.indexer.k_norm",
+      "model.layers.60.self_attn.indexer.k_norm.bias",
+      "model.layers.60.self_attn.indexers_proj",
+      "model.layers.60.self_attn.kv_a_layernorm",
+      "model.layers.60.self_attn.q_a_layernorm",
+      "model.layers.61.input_layernorm",
+      "model.layers.61.mlp.gate",
+      "model.layers.61.mlp.gate.e_score_correction_bias",
+      "model.layers.61.post_attention_layernorm",
+      "model.layers.61.self_attn.indexer.k_norm",
+      "model.layers.61.self_attn.indexer.k_norm.bias",
+      "model.layers.61.self_attn.indexers_proj",
+      "model.layers.61.self_attn.kv_a_layernorm",
+      "model.layers.61.self_attn.q_a_layernorm",
+      "model.layers.62.input_layernorm",
+      "model.layers.62.mlp.gate",
+      "model.layers.62.mlp.gate.e_score_correction_bias",
+      "model.layers.62.post_attention_layernorm",
+      "model.layers.62.self_attn.indexer.k_norm",
+      "model.layers.62.self_attn.indexer.k_norm.bias",
+      "model.layers.62.self_attn.indexers_proj",
+      "model.layers.62.self_attn.kv_a_layernorm",
+      "model.layers.62.self_attn.q_a_layernorm",
+      "model.layers.63.input_layernorm",
+      "model.layers.63.mlp.gate",
+      "model.layers.63.mlp.gate.e_score_correction_bias",
+      "model.layers.63.post_attention_layernorm",
+      "model.layers.63.self_attn.indexer.k_norm",
+      "model.layers.63.self_attn.indexer.k_norm.bias",
+      "model.layers.63.self_attn.indexers_proj",
+      "model.layers.63.self_attn.kv_a_layernorm",
+      "model.layers.63.self_attn.q_a_layernorm",
+      "model.layers.64.input_layernorm",
+      "model.layers.64.mlp.gate",
+      "model.layers.64.mlp.gate.e_score_correction_bias",
+      "model.layers.64.post_attention_layernorm",
+      "model.layers.64.self_attn.indexer.k_norm",
+      "model.layers.64.self_attn.indexer.k_norm.bias",
+      "model.layers.64.self_attn.indexers_proj",
+      "model.layers.64.self_attn.kv_a_layernorm",
+      "model.layers.64.self_attn.q_a_layernorm",
+      "model.layers.65.input_layernorm",
+      "model.layers.65.mlp.gate",
+      "model.layers.65.mlp.gate.e_score_correction_bias",
+      "model.layers.65.post_attention_layernorm",
+      "model.layers.65.self_attn.indexer.k_norm",
+      "model.layers.65.self_attn.indexer.k_norm.bias",
+      "model.layers.65.self_attn.indexers_proj",
+      "model.layers.65.self_attn.kv_a_layernorm",
+      "model.layers.65.self_attn.q_a_layernorm",
+      "model.layers.66.input_layernorm",
+      "model.layers.66.mlp.gate",
+      "model.layers.66.mlp.gate.e_score_correction_bias",
+      "model.layers.66.post_attention_layernorm",
+      "model.layers.66.self_attn.indexer.k_norm",
+      "model.layers.66.self_attn.indexer.k_norm.bias",
+      "model.layers.66.self_attn.indexers_proj",
+      "model.layers.66.self_attn.kv_a_layernorm",
+      "model.layers.66.self_attn.q_a_layernorm",
+      "model.layers.67.input_layernorm",
+      "model.layers.67.mlp.gate",
+      "model.layers.67.mlp.gate.e_score_correction_bias",
+      "model.layers.67.post_attention_layernorm",
+      "model.layers.67.self_attn.indexer.k_norm",
+      "model.layers.67.self_attn.indexer.k_norm.bias",
+      "model.layers.67.self_attn.indexers_proj",
+      "model.layers.67.self_attn.kv_a_layernorm",
+      "model.layers.67.self_attn.q_a_layernorm",
+      "model.layers.68.input_layernorm",
+      "model.layers.68.mlp.gate",
+      "model.layers.68.mlp.gate.e_score_correction_bias",
+      "model.layers.68.post_attention_layernorm",
+      "model.layers.68.self_attn.indexer.k_norm",
+      "model.layers.68.self_attn.indexer.k_norm.bias",
+      "model.layers.68.self_attn.indexers_proj",
+      "model.layers.68.self_attn.kv_a_layernorm",
+      "model.layers.68.self_attn.q_a_layernorm",
+      "model.layers.69.input_layernorm",
+      "model.layers.69.mlp.gate",
+      "model.layers.69.mlp.gate.e_score_correction_bias",
+      "model.layers.69.post_attention_layernorm",
+      "model.layers.69.self_attn.indexer.k_norm",
+      "model.layers.69.self_attn.indexer.k_norm.bias",
+      "model.layers.69.self_attn.indexers_proj",
+      "model.layers.69.self_attn.kv_a_layernorm",
+      "model.layers.69.self_attn.q_a_layernorm",
+      "model.layers.70.input_layernorm",
+      "model.layers.70.mlp.gate",
+      "model.layers.70.mlp.gate.e_score_correction_bias",
+      "model.layers.70.post_attention_layernorm",
+      "model.layers.70.self_attn.indexer.k_norm",
+      "model.layers.70.self_attn.indexer.k_norm.bias",
+      "model.layers.70.self_attn.indexers_proj",
+      "model.layers.70.self_attn.kv_a_layernorm",
+      "model.layers.70.self_attn.q_a_layernorm",
+      "model.layers.71.input_layernorm",
+      "model.layers.71.mlp.gate",
+      "model.layers.71.mlp.gate.e_score_correction_bias",
+      "model.layers.71.post_attention_layernorm",
+      "model.layers.71.self_attn.indexer.k_norm",
+      "model.layers.71.self_attn.indexer.k_norm.bias",
+      "model.layers.71.self_attn.indexers_proj",
+      "model.layers.71.self_attn.kv_a_layernorm",
+      "model.layers.71.self_attn.q_a_layernorm",
+      "model.layers.72.input_layernorm",
+      "model.layers.72.mlp.gate",
+      "model.layers.72.mlp.gate.e_score_correction_bias",
+      "model.layers.72.post_attention_layernorm",
+      "model.layers.72.self_attn.indexer.k_norm",
+      "model.layers.72.self_attn.indexer.k_norm.bias",
+      "model.layers.72.self_attn.indexers_proj",
+      "model.layers.72.self_attn.kv_a_layernorm",
+      "model.layers.72.self_attn.q_a_layernorm",
+      "model.layers.73.input_layernorm",
+      "model.layers.73.mlp.gate",
+      "model.layers.73.mlp.gate.e_score_correction_bias",
+      "model.layers.73.post_attention_layernorm",
+      "model.layers.73.self_attn.indexer.k_norm",
+      "model.layers.73.self_attn.indexer.k_norm.bias",
+      "model.layers.73.self_attn.indexers_proj",
+      "model.layers.73.self_attn.kv_a_layernorm",
+      "model.layers.73.self_attn.q_a_layernorm",
+      "model.layers.74.input_layernorm",
+      "model.layers.74.mlp.gate",
+      "model.layers.74.mlp.gate.e_score_correction_bias",
+      "model.layers.74.post_attention_layernorm",
+      "model.layers.74.self_attn.indexer.k_norm",
+      "model.layers.74.self_attn.indexer.k_norm.bias",
+      "model.layers.74.self_attn.indexers_proj",
+      "model.layers.74.self_attn.kv_a_layernorm",
+      "model.layers.74.self_attn.q_a_layernorm",
+      "model.layers.75.input_layernorm",
+      "model.layers.75.mlp.gate",
+      "model.layers.75.mlp.gate.e_score_correction_bias",
+      "model.layers.75.post_attention_layernorm",
+      "model.layers.75.self_attn.indexer.k_norm",
+      "model.layers.75.self_attn.indexer.k_norm.bias",
+      "model.layers.75.self_attn.indexers_proj",
+      "model.layers.75.self_attn.kv_a_layernorm",
+      "model.layers.75.self_attn.q_a_layernorm",
+      "model.layers.76.input_layernorm",
+      "model.layers.76.mlp.gate",
+      "model.layers.76.mlp.gate.e_score_correction_bias",
+      "model.layers.76.post_attention_layernorm",
+      "model.layers.76.self_attn.indexer.k_norm",
+      "model.layers.76.self_attn.indexer.k_norm.bias",
+      "model.layers.76.self_attn.indexers_proj",
+      "model.layers.76.self_attn.kv_a_layernorm",
+      "model.layers.76.self_attn.q_a_layernorm",
+      "model.layers.77.input_layernorm",
+      "model.layers.77.mlp.gate",
+      "model.layers.77.mlp.gate.e_score_correction_bias",
+      "model.layers.77.post_attention_layernorm",
+      "model.layers.77.self_attn.indexer.k_norm",
+      "model.layers.77.self_attn.indexer.k_norm.bias",
+      "model.layers.77.self_attn.indexers_proj",
+      "model.layers.77.self_attn.kv_a_layernorm",
+      "model.layers.77.self_attn.q_a_layernorm",
+      "model.layers.78.eh_proj",
+      "model.layers.78.enorm",
+      "model.layers.78.hnorm",
+      "model.layers.78.input_layernorm",
+      "model.layers.78.mlp.gate",
+      "model.layers.78.mlp.gate.e_score_correction_bias",
+      "model.layers.78.post_attention_layernorm",
+      "model.layers.78.self_attn.indexer.k_norm",
+      "model.layers.78.self_attn.indexer.k_norm.bias",
+      "model.layers.78.self_attn.indexers_proj",
+      "model.layers.78.self_attn.kv_a_layernorm",
+      "model.layers.78.self_attn.q_a_layernorm",
+      "model.layers.78.shared_head.norm",
+      "model.norm"
+    ],
+    "quant_method": "fp8",
+    "weight_block_size": [
+      128,
+      128
+    ]
+  },
+  "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": "5.3.0.dev0",
+  "use_cache": true,
+  "v_head_dim": 256,
+  "vocab_size": 154880
+}

generation_config.json

@@ -0,0 +1,13 @@
+{
+  "_from_model_config": true,
+  "do_sample": true,
+  "eos_token_id": [
+    154820,
+    154827,
+    154829
+  ],
+  "pad_token_id": 154820,
+  "temperature": 1.0,
+  "top_p": 0.95,
+  "transformers_version": "5.3.0.dev0"
+}

model-00001-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:416021f585df721f620359a57d1d9e3175d833bb98ecefdc9a395700b95d404c
+size 47759253472

model-00002-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:733e45e7a88ce29c33c8620c8ca4113a0904ecd3afa6dfa2016e9c0d3b08047d
+size 49409000576

model-00003-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f9373370f7993e066da2b8bf65b8506b8957691294b8e7ecb357d3dd82bae39
+size 49409004760

model-00004-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d94625e1d4a1b2c0365c9288f8ef8a6513439502b67e8de9940a6f473803b49
+size 49409004760

model-00005-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99f0e114376499c9ca927c37a5b1170989e782b89d51ffc8f91a0763b48b3745
+size 49409004760

model-00006-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0cba6ff4f1b23dea839b67c2828c4894a12d8eebaf5498a2d9f9181ea2c27d9e
+size 49409004760

model-00007-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca35ba9d4ec228a47a3a34c4d88b84e7ddd101f4ba6ae8b8190d243b06b8be88
+size 49409004760

model-00008-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14b10760034afae4684d2dc4555d4f0b176055c9dc963e09cf739f17319e05ff
+size 49409004760

model-00009-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73ff38608e6e3ded29149089f3e76de32d2d879a1b3819135b1fc3d6d308a0df
+size 49409004760

model-00010-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:225655e2e240a4b17ea6b09a66ae74e4bbb800b79e0cc5127cc2868b75ea2c85
+size 49409004760

model-00011-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1dec06942660a0e1f9c257f01c121c2b53cdbb597f01f73fe67d8600878abfe2
+size 49409004760

model-00012-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96d4016626e94d9190669cbf674b5e89574aec3871223327e5db2d3f28bf5c56
+size 49409004760

model-00013-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73bd4342291705287e57abe9adfb7c608193c1e2f0531c0235e15386c445ca4a
+size 49409004760

model-00014-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e15507743d6db1be01fa9328cd2f0c7ed37e9dd9e870ec1a568ca3f000a45a9b
+size 49409004760

model-00015-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bc919a81816214f3f5befc58af054336a9fa6eb328b33db7a228e976c515f20
+size 49409004760

model-00016-of-00016.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a025b72eee610d77ca0d8ff54a6ace589aec7189f186a55643942bdd4cdd26d9
+size 6659721496

model.safetensors.index.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7884e881184fa865c89a622f56a5435c7122bf8e4329c448573fa0f87426c3ff
+size 11245948

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"
+}