Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+model.safetensors.index.json filter=lfs diff=lfs merge=lfs -text

chat_template.jinja

@@ -0,0 +1,46 @@
+{%- if tools -%}
+  <|im_system|>tool_declare<|im_middle|>
+  # Tools
+  {{ tools | tojson }}<|im_end|>
+{%- endif -%}
+{%- for message in messages -%}
+  {%- if loop.first and messages[0]['role'] != 'system' -%}
+    <|im_system|>system<|im_middle|>You are Kimi, an AI assistant created by Moonshot AI.<|im_end|>
+  {%- endif -%}
+  
+  {%- set role_name =  message.get('name') or  message['role'] -%}
+  {%- if message['role'] == 'user' -%}
+    <|im_user|>{{role_name}}<|im_middle|>
+  {%- elif message['role'] == 'assistant' -%}
+    <|im_assistant|>{{role_name}}<|im_middle|>
+  {%- else -%}
+    <|im_system|>{{role_name}}<|im_middle|>
+  {%- endif -%}
+
+  {%- if message['role'] == 'assistant' and message.get('tool_calls') -%}
+    {%- if message['content'] -%}{{ message['content'] }}{%- endif -%}
+    <|tool_calls_section_begin|>
+    {%- for tool_call in message['tool_calls'] -%}
+      {%- set formatted_id = tool_call['id'] -%}
+      <|tool_call_begin|>{{ formatted_id }}<|tool_call_argument_begin|>{% if tool_call['function']['arguments'] is string %}{{ tool_call['function']['arguments'] }}{% else %}{{ tool_call['function']['arguments'] | tojson }}{% endif %}<|tool_call_end|>
+    {%- endfor -%}
+    <|tool_calls_section_end|>
+  {%- elif message['role'] == 'tool' -%}
+    ## Return of {{ message.tool_call_id }}
+    {{ message['content'] }}
+  {%- elif message['content'] is string -%}
+    {{ message['content'] }}
+  {%- elif message['content'] is not none -%}
+    {% for content in message['content'] -%}
+      {% if content['type'] == 'image' or 'image' in content or 'image_url' in content -%}
+        <|media_start|>image<|media_content|><|media_pad|><|media_end|>
+      {% else -%}
+        {{ content['text'] }}
+      {%- endif -%}
+    {%- endfor -%}
+  {%- endif -%}
+  <|im_end|>
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+  <|im_assistant|>assistant<|im_middle|>
+{%- endif -%}

config.json

@@ -0,0 +1,1480 @@
+{
+  "_name_or_path": "/mnt/weka/home/ggb/mergekit/kimi-k2-merged",
+  "add_cross_attention": false,
+  "architectures": [
+    "DeepseekV3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_deepseek.DeepseekV3Config",
+    "AutoModel": "modeling_deepseek.DeepseekV3Model",
+    "AutoModelForCausalLM": "modeling_deepseek.DeepseekV3ForCausalLM"
+  },
+  "aux_loss_alpha": 0.001,
+  "bad_words_ids": null,
+  "begin_suppress_tokens": null,
+  "bos_token_id": 163584,
+  "chunk_size_feed_forward": 0,
+  "cross_attention_hidden_size": null,
+  "decoder_start_token_id": null,
+  "diversity_penalty": 0.0,
+  "do_sample": false,
+  "early_stopping": false,
+  "encoder_no_repeat_ngram_size": 0,
+  "eos_token_id": 163585,
+  "ep_size": 1,
+  "exponential_decay_length_penalty": null,
+  "finetuning_task": null,
+  "first_k_dense_replace": 1,
+  "forced_bos_token_id": null,
+  "forced_eos_token_id": null,
+  "head_dim": 64,
+  "hidden_act": "silu",
+  "hidden_size": 7168,
+  "id2label": {
+    "0": "LABEL_0",
+    "1": "LABEL_1"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 18432,
+  "is_decoder": false,
+  "is_encoder_decoder": false,
+  "k3_headstack_recipe": {
+    "attention": "q_b_proj and kv_b_proj row-concat base plus mapped instruct branch; o_proj col-concat with per-variant scale",
+    "attention_a_proj": "q_a_proj and kv_a_proj_with_mqa rowwise SLERP base plus mapped instruct branch",
+    "attention_a_proj_slerp_dot_threshold": 0.9995,
+    "attention_a_proj_slerp_t": 0.5,
+    "o_proj_scale": 0.5,
+    "shared_experts": "gate/up row-concat and down col-concat base plus mapped instruct branch"
+  },
+  "kv_lora_rank": 512,
+  "label2id": {
+    "LABEL_0": 0,
+    "LABEL_1": 1
+  },
+  "length_penalty": 1.0,
+  "max_length": 20,
+  "max_position_embeddings": 131072,
+  "min_length": 0,
+  "model_type": "deepseek_v3",
+  "moe_intermediate_size": 2048,
+  "moe_layer_freq": 1,
+  "n_group": 1,
+  "n_routed_experts": 768,
+  "n_shared_experts": 2,
+  "no_repeat_ngram_size": 0,
+  "norm_topk_prob": true,
+  "num_attention_heads": 128,
+  "num_beam_groups": 1,
+  "num_beams": 1,
+  "num_experts_per_tok": 16,
+  "num_hidden_layers": 103,
+  "num_key_value_heads": 128,
+  "num_nextn_predict_layers": 0,
+  "num_return_sequences": 1,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "output_scores": false,
+  "pad_token_id": null,
+  "prefix": null,
+  "pretraining_tp": 1,
+  "problem_type": null,
+  "pruned_heads": {},
+  "q_lora_rank": 1536,
+  "qk_head_dim": 192,
+  "qk_nope_head_dim": 128,
+  "qk_rope_head_dim": 64,
+  "quantization_config": {
+    "activation_scheme": "dynamic",
+    "fmt": "e4m3",
+    "ignored_layers": [
+      "lm_head",
+      "model.embed_tokens",
+      "model.layers.0.input_layernorm",
+      "model.layers.0.mlp.down_proj",
+      "model.layers.0.mlp.gate_proj",
+      "model.layers.0.mlp.up_proj",
+      "model.layers.0.post_attention_layernorm",
+      "model.layers.0.self_attn.kv_a_layernorm",
+      "model.layers.0.self_attn.kv_a_proj_with_mqa",
+      "model.layers.0.self_attn.kv_b_proj",
+      "model.layers.0.self_attn.o_proj",
+      "model.layers.0.self_attn.q_a_layernorm",
+      "model.layers.0.self_attn.q_a_proj",
+      "model.layers.0.self_attn.q_b_proj",
+      "model.layers.1.input_layernorm",
+      "model.layers.1.mlp.gate",
+      "model.layers.1.mlp.shared_experts.down_proj",
+      "model.layers.1.mlp.shared_experts.gate_proj",
+      "model.layers.1.mlp.shared_experts.up_proj",
+      "model.layers.1.post_attention_layernorm",
+      "model.layers.1.self_attn.kv_a_layernorm",
+      "model.layers.1.self_attn.kv_a_proj_with_mqa",
+      "model.layers.1.self_attn.kv_b_proj",
+      "model.layers.1.self_attn.o_proj",
+      "model.layers.1.self_attn.q_a_layernorm",
+      "model.layers.1.self_attn.q_a_proj",
+      "model.layers.1.self_attn.q_b_proj",
+      "model.layers.10.input_layernorm",
+      "model.layers.10.mlp.gate",
+      "model.layers.10.mlp.shared_experts.down_proj",
+      "model.layers.10.mlp.shared_experts.gate_proj",
+      "model.layers.10.mlp.shared_experts.up_proj",
+      "model.layers.10.post_attention_layernorm",
+      "model.layers.10.self_attn.kv_a_layernorm",
+      "model.layers.10.self_attn.kv_a_proj_with_mqa",
+      "model.layers.10.self_attn.kv_b_proj",
+      "model.layers.10.self_attn.o_proj",
+      "model.layers.10.self_attn.q_a_layernorm",
+      "model.layers.10.self_attn.q_a_proj",
+      "model.layers.10.self_attn.q_b_proj",
+      "model.layers.100.input_layernorm",
+      "model.layers.100.mlp.gate",
+      "model.layers.100.mlp.shared_experts.down_proj",
+      "model.layers.100.mlp.shared_experts.gate_proj",
+      "model.layers.100.mlp.shared_experts.up_proj",
+      "model.layers.100.post_attention_layernorm",
+      "model.layers.100.self_attn.kv_a_layernorm",
+      "model.layers.100.self_attn.kv_a_proj_with_mqa",
+      "model.layers.100.self_attn.kv_b_proj",
+      "model.layers.100.self_attn.o_proj",
+      "model.layers.100.self_attn.q_a_layernorm",
+      "model.layers.100.self_attn.q_a_proj",
+      "model.layers.100.self_attn.q_b_proj",
+      "model.layers.101.input_layernorm",
+      "model.layers.101.mlp.gate",
+      "model.layers.101.mlp.shared_experts.down_proj",
+      "model.layers.101.mlp.shared_experts.gate_proj",
+      "model.layers.101.mlp.shared_experts.up_proj",
+      "model.layers.101.post_attention_layernorm",
+      "model.layers.101.self_attn.kv_a_layernorm",
+      "model.layers.101.self_attn.kv_a_proj_with_mqa",
+      "model.layers.101.self_attn.kv_b_proj",
+      "model.layers.101.self_attn.o_proj",
+      "model.layers.101.self_attn.q_a_layernorm",
+      "model.layers.101.self_attn.q_a_proj",
+      "model.layers.101.self_attn.q_b_proj",
+      "model.layers.102.input_layernorm",
+      "model.layers.102.mlp.gate",
+      "model.layers.102.mlp.shared_experts.down_proj",
+      "model.layers.102.mlp.shared_experts.gate_proj",
+      "model.layers.102.mlp.shared_experts.up_proj",
+      "model.layers.102.post_attention_layernorm",
+      "model.layers.102.self_attn.kv_a_layernorm",
+      "model.layers.102.self_attn.kv_a_proj_with_mqa",
+      "model.layers.102.self_attn.kv_b_proj",
+      "model.layers.102.self_attn.o_proj",
+      "model.layers.102.self_attn.q_a_layernorm",
+      "model.layers.102.self_attn.q_a_proj",
+      "model.layers.102.self_attn.q_b_proj",
+      "model.layers.11.input_layernorm",
+      "model.layers.11.mlp.gate",
+      "model.layers.11.mlp.shared_experts.down_proj",
+      "model.layers.11.mlp.shared_experts.gate_proj",
+      "model.layers.11.mlp.shared_experts.up_proj",
+      "model.layers.11.post_attention_layernorm",
+      "model.layers.11.self_attn.kv_a_layernorm",
+      "model.layers.11.self_attn.kv_a_proj_with_mqa",
+      "model.layers.11.self_attn.kv_b_proj",
+      "model.layers.11.self_attn.o_proj",
+      "model.layers.11.self_attn.q_a_layernorm",
+      "model.layers.11.self_attn.q_a_proj",
+      "model.layers.11.self_attn.q_b_proj",
+      "model.layers.12.input_layernorm",
+      "model.layers.12.mlp.gate",
+      "model.layers.12.mlp.shared_experts.down_proj",
+      "model.layers.12.mlp.shared_experts.gate_proj",
+      "model.layers.12.mlp.shared_experts.up_proj",
+      "model.layers.12.post_attention_layernorm",
+      "model.layers.12.self_attn.kv_a_layernorm",
+      "model.layers.12.self_attn.kv_a_proj_with_mqa",
+      "model.layers.12.self_attn.kv_b_proj",
+      "model.layers.12.self_attn.o_proj",
+      "model.layers.12.self_attn.q_a_layernorm",
+      "model.layers.12.self_attn.q_a_proj",
+      "model.layers.12.self_attn.q_b_proj",
+      "model.layers.13.input_layernorm",
+      "model.layers.13.mlp.gate",
+      "model.layers.13.mlp.shared_experts.down_proj",
+      "model.layers.13.mlp.shared_experts.gate_proj",
+      "model.layers.13.mlp.shared_experts.up_proj",
+      "model.layers.13.post_attention_layernorm",
+      "model.layers.13.self_attn.kv_a_layernorm",
+      "model.layers.13.self_attn.kv_a_proj_with_mqa",
+      "model.layers.13.self_attn.kv_b_proj",
+      "model.layers.13.self_attn.o_proj",
+      "model.layers.13.self_attn.q_a_layernorm",
+      "model.layers.13.self_attn.q_a_proj",
+      "model.layers.13.self_attn.q_b_proj",
+      "model.layers.14.input_layernorm",
+      "model.layers.14.mlp.gate",
+      "model.layers.14.mlp.shared_experts.down_proj",
+      "model.layers.14.mlp.shared_experts.gate_proj",
+      "model.layers.14.mlp.shared_experts.up_proj",
+      "model.layers.14.post_attention_layernorm",
+      "model.layers.14.self_attn.kv_a_layernorm",
+      "model.layers.14.self_attn.kv_a_proj_with_mqa",
+      "model.layers.14.self_attn.kv_b_proj",
+      "model.layers.14.self_attn.o_proj",
+      "model.layers.14.self_attn.q_a_layernorm",
+      "model.layers.14.self_attn.q_a_proj",
+      "model.layers.14.self_attn.q_b_proj",
+      "model.layers.15.input_layernorm",
+      "model.layers.15.mlp.gate",
+      "model.layers.15.mlp.shared_experts.down_proj",
+      "model.layers.15.mlp.shared_experts.gate_proj",
+      "model.layers.15.mlp.shared_experts.up_proj",
+      "model.layers.15.post_attention_layernorm",
+      "model.layers.15.self_attn.kv_a_layernorm",
+      "model.layers.15.self_attn.kv_a_proj_with_mqa",
+      "model.layers.15.self_attn.kv_b_proj",
+      "model.layers.15.self_attn.o_proj",
+      "model.layers.15.self_attn.q_a_layernorm",
+      "model.layers.15.self_attn.q_a_proj",
+      "model.layers.15.self_attn.q_b_proj",
+      "model.layers.16.input_layernorm",
+      "model.layers.16.mlp.gate",
+      "model.layers.16.mlp.shared_experts.down_proj",
+      "model.layers.16.mlp.shared_experts.gate_proj",
+      "model.layers.16.mlp.shared_experts.up_proj",
+      "model.layers.16.post_attention_layernorm",
+      "model.layers.16.self_attn.kv_a_layernorm",
+      "model.layers.16.self_attn.kv_a_proj_with_mqa",
+      "model.layers.16.self_attn.kv_b_proj",
+      "model.layers.16.self_attn.o_proj",
+      "model.layers.16.self_attn.q_a_layernorm",
+      "model.layers.16.self_attn.q_a_proj",
+      "model.layers.16.self_attn.q_b_proj",
+      "model.layers.17.input_layernorm",
+      "model.layers.17.mlp.gate",
+      "model.layers.17.mlp.shared_experts.down_proj",
+      "model.layers.17.mlp.shared_experts.gate_proj",
+      "model.layers.17.mlp.shared_experts.up_proj",
+      "model.layers.17.post_attention_layernorm",
+      "model.layers.17.self_attn.kv_a_layernorm",
+      "model.layers.17.self_attn.kv_a_proj_with_mqa",
+      "model.layers.17.self_attn.kv_b_proj",
+      "model.layers.17.self_attn.o_proj",
+      "model.layers.17.self_attn.q_a_layernorm",
+      "model.layers.17.self_attn.q_a_proj",
+      "model.layers.17.self_attn.q_b_proj",
+      "model.layers.18.input_layernorm",
+      "model.layers.18.mlp.gate",
+      "model.layers.18.mlp.shared_experts.down_proj",
+      "model.layers.18.mlp.shared_experts.gate_proj",
+      "model.layers.18.mlp.shared_experts.up_proj",
+      "model.layers.18.post_attention_layernorm",
+      "model.layers.18.self_attn.kv_a_layernorm",
+      "model.layers.18.self_attn.kv_a_proj_with_mqa",
+      "model.layers.18.self_attn.kv_b_proj",
+      "model.layers.18.self_attn.o_proj",
+      "model.layers.18.self_attn.q_a_layernorm",
+      "model.layers.18.self_attn.q_a_proj",
+      "model.layers.18.self_attn.q_b_proj",
+      "model.layers.19.input_layernorm",
+      "model.layers.19.mlp.gate",
+      "model.layers.19.mlp.shared_experts.down_proj",
+      "model.layers.19.mlp.shared_experts.gate_proj",
+      "model.layers.19.mlp.shared_experts.up_proj",
+      "model.layers.19.post_attention_layernorm",
+      "model.layers.19.self_attn.kv_a_layernorm",
+      "model.layers.19.self_attn.kv_a_proj_with_mqa",
+      "model.layers.19.self_attn.kv_b_proj",
+      "model.layers.19.self_attn.o_proj",
+      "model.layers.19.self_attn.q_a_layernorm",
+      "model.layers.19.self_attn.q_a_proj",
+      "model.layers.19.self_attn.q_b_proj",
+      "model.layers.2.input_layernorm",
+      "model.layers.2.mlp.gate",
+      "model.layers.2.mlp.shared_experts.down_proj",
+      "model.layers.2.mlp.shared_experts.gate_proj",
+      "model.layers.2.mlp.shared_experts.up_proj",
+      "model.layers.2.post_attention_layernorm",
+      "model.layers.2.self_attn.kv_a_layernorm",
+      "model.layers.2.self_attn.kv_a_proj_with_mqa",
+      "model.layers.2.self_attn.kv_b_proj",
+      "model.layers.2.self_attn.o_proj",
+      "model.layers.2.self_attn.q_a_layernorm",
+      "model.layers.2.self_attn.q_a_proj",
+      "model.layers.2.self_attn.q_b_proj",
+      "model.layers.20.input_layernorm",
+      "model.layers.20.mlp.gate",
+      "model.layers.20.mlp.shared_experts.down_proj",
+      "model.layers.20.mlp.shared_experts.gate_proj",
+      "model.layers.20.mlp.shared_experts.up_proj",
+      "model.layers.20.post_attention_layernorm",
+      "model.layers.20.self_attn.kv_a_layernorm",
+      "model.layers.20.self_attn.kv_a_proj_with_mqa",
+      "model.layers.20.self_attn.kv_b_proj",
+      "model.layers.20.self_attn.o_proj",
+      "model.layers.20.self_attn.q_a_layernorm",
+      "model.layers.20.self_attn.q_a_proj",
+      "model.layers.20.self_attn.q_b_proj",
+      "model.layers.21.input_layernorm",
+      "model.layers.21.mlp.gate",
+      "model.layers.21.mlp.shared_experts.down_proj",
+      "model.layers.21.mlp.shared_experts.gate_proj",
+      "model.layers.21.mlp.shared_experts.up_proj",
+      "model.layers.21.post_attention_layernorm",
+      "model.layers.21.self_attn.kv_a_layernorm",
+      "model.layers.21.self_attn.kv_a_proj_with_mqa",
+      "model.layers.21.self_attn.kv_b_proj",
+      "model.layers.21.self_attn.o_proj",
+      "model.layers.21.self_attn.q_a_layernorm",
+      "model.layers.21.self_attn.q_a_proj",
+      "model.layers.21.self_attn.q_b_proj",
+      "model.layers.22.input_layernorm",
+      "model.layers.22.mlp.gate",
+      "model.layers.22.mlp.shared_experts.down_proj",
+      "model.layers.22.mlp.shared_experts.gate_proj",
+      "model.layers.22.mlp.shared_experts.up_proj",
+      "model.layers.22.post_attention_layernorm",
+      "model.layers.22.self_attn.kv_a_layernorm",
+      "model.layers.22.self_attn.kv_a_proj_with_mqa",
+      "model.layers.22.self_attn.kv_b_proj",
+      "model.layers.22.self_attn.o_proj",
+      "model.layers.22.self_attn.q_a_layernorm",
+      "model.layers.22.self_attn.q_a_proj",
+      "model.layers.22.self_attn.q_b_proj",
+      "model.layers.23.input_layernorm",
+      "model.layers.23.mlp.gate",
+      "model.layers.23.mlp.shared_experts.down_proj",
+      "model.layers.23.mlp.shared_experts.gate_proj",
+      "model.layers.23.mlp.shared_experts.up_proj",
+      "model.layers.23.post_attention_layernorm",
+      "model.layers.23.self_attn.kv_a_layernorm",
+      "model.layers.23.self_attn.kv_a_proj_with_mqa",
+      "model.layers.23.self_attn.kv_b_proj",
+      "model.layers.23.self_attn.o_proj",
+      "model.layers.23.self_attn.q_a_layernorm",
+      "model.layers.23.self_attn.q_a_proj",
+      "model.layers.23.self_attn.q_b_proj",
+      "model.layers.24.input_layernorm",
+      "model.layers.24.mlp.gate",
+      "model.layers.24.mlp.shared_experts.down_proj",
+      "model.layers.24.mlp.shared_experts.gate_proj",
+      "model.layers.24.mlp.shared_experts.up_proj",
+      "model.layers.24.post_attention_layernorm",
+      "model.layers.24.self_attn.kv_a_layernorm",
+      "model.layers.24.self_attn.kv_a_proj_with_mqa",
+      "model.layers.24.self_attn.kv_b_proj",
+      "model.layers.24.self_attn.o_proj",
+      "model.layers.24.self_attn.q_a_layernorm",
+      "model.layers.24.self_attn.q_a_proj",
+      "model.layers.24.self_attn.q_b_proj",
+      "model.layers.25.input_layernorm",
+      "model.layers.25.mlp.gate",
+      "model.layers.25.mlp.shared_experts.down_proj",
+      "model.layers.25.mlp.shared_experts.gate_proj",
+      "model.layers.25.mlp.shared_experts.up_proj",
+      "model.layers.25.post_attention_layernorm",
+      "model.layers.25.self_attn.kv_a_layernorm",
+      "model.layers.25.self_attn.kv_a_proj_with_mqa",
+      "model.layers.25.self_attn.kv_b_proj",
+      "model.layers.25.self_attn.o_proj",
+      "model.layers.25.self_attn.q_a_layernorm",
+      "model.layers.25.self_attn.q_a_proj",
+      "model.layers.25.self_attn.q_b_proj",
+      "model.layers.26.input_layernorm",
+      "model.layers.26.mlp.gate",
+      "model.layers.26.mlp.shared_experts.down_proj",
+      "model.layers.26.mlp.shared_experts.gate_proj",
+      "model.layers.26.mlp.shared_experts.up_proj",
+      "model.layers.26.post_attention_layernorm",
+      "model.layers.26.self_attn.kv_a_layernorm",
+      "model.layers.26.self_attn.kv_a_proj_with_mqa",
+      "model.layers.26.self_attn.kv_b_proj",
+      "model.layers.26.self_attn.o_proj",
+      "model.layers.26.self_attn.q_a_layernorm",
+      "model.layers.26.self_attn.q_a_proj",
+      "model.layers.26.self_attn.q_b_proj",
+      "model.layers.27.input_layernorm",
+      "model.layers.27.mlp.gate",
+      "model.layers.27.mlp.shared_experts.down_proj",
+      "model.layers.27.mlp.shared_experts.gate_proj",
+      "model.layers.27.mlp.shared_experts.up_proj",
+      "model.layers.27.post_attention_layernorm",
+      "model.layers.27.self_attn.kv_a_layernorm",
+      "model.layers.27.self_attn.kv_a_proj_with_mqa",
+      "model.layers.27.self_attn.kv_b_proj",
+      "model.layers.27.self_attn.o_proj",
+      "model.layers.27.self_attn.q_a_layernorm",
+      "model.layers.27.self_attn.q_a_proj",
+      "model.layers.27.self_attn.q_b_proj",
+      "model.layers.28.input_layernorm",
+      "model.layers.28.mlp.gate",
+      "model.layers.28.mlp.shared_experts.down_proj",
+      "model.layers.28.mlp.shared_experts.gate_proj",
+      "model.layers.28.mlp.shared_experts.up_proj",
+      "model.layers.28.post_attention_layernorm",
+      "model.layers.28.self_attn.kv_a_layernorm",
+      "model.layers.28.self_attn.kv_a_proj_with_mqa",
+      "model.layers.28.self_attn.kv_b_proj",
+      "model.layers.28.self_attn.o_proj",
+      "model.layers.28.self_attn.q_a_layernorm",
+      "model.layers.28.self_attn.q_a_proj",
+      "model.layers.28.self_attn.q_b_proj",
+      "model.layers.29.input_layernorm",
+      "model.layers.29.mlp.gate",
+      "model.layers.29.mlp.shared_experts.down_proj",
+      "model.layers.29.mlp.shared_experts.gate_proj",
+      "model.layers.29.mlp.shared_experts.up_proj",
+      "model.layers.29.post_attention_layernorm",
+      "model.layers.29.self_attn.kv_a_layernorm",
+      "model.layers.29.self_attn.kv_a_proj_with_mqa",
+      "model.layers.29.self_attn.kv_b_proj",
+      "model.layers.29.self_attn.o_proj",
+      "model.layers.29.self_attn.q_a_layernorm",
+      "model.layers.29.self_attn.q_a_proj",
+      "model.layers.29.self_attn.q_b_proj",
+      "model.layers.3.input_layernorm",
+      "model.layers.3.mlp.gate",
+      "model.layers.3.mlp.shared_experts.down_proj",
+      "model.layers.3.mlp.shared_experts.gate_proj",
+      "model.layers.3.mlp.shared_experts.up_proj",
+      "model.layers.3.post_attention_layernorm",
+      "model.layers.3.self_attn.kv_a_layernorm",
+      "model.layers.3.self_attn.kv_a_proj_with_mqa",
+      "model.layers.3.self_attn.kv_b_proj",
+      "model.layers.3.self_attn.o_proj",
+      "model.layers.3.self_attn.q_a_layernorm",
+      "model.layers.3.self_attn.q_a_proj",
+      "model.layers.3.self_attn.q_b_proj",
+      "model.layers.30.input_layernorm",
+      "model.layers.30.mlp.gate",
+      "model.layers.30.mlp.shared_experts.down_proj",
+      "model.layers.30.mlp.shared_experts.gate_proj",
+      "model.layers.30.mlp.shared_experts.up_proj",
+      "model.layers.30.post_attention_layernorm",
+      "model.layers.30.self_attn.kv_a_layernorm",
+      "model.layers.30.self_attn.kv_a_proj_with_mqa",
+      "model.layers.30.self_attn.kv_b_proj",
+      "model.layers.30.self_attn.o_proj",
+      "model.layers.30.self_attn.q_a_layernorm",
+      "model.layers.30.self_attn.q_a_proj",
+      "model.layers.30.self_attn.q_b_proj",
+      "model.layers.31.input_layernorm",
+      "model.layers.31.mlp.gate",
+      "model.layers.31.mlp.shared_experts.down_proj",
+      "model.layers.31.mlp.shared_experts.gate_proj",
+      "model.layers.31.mlp.shared_experts.up_proj",
+      "model.layers.31.post_attention_layernorm",
+      "model.layers.31.self_attn.kv_a_layernorm",
+      "model.layers.31.self_attn.kv_a_proj_with_mqa",
+      "model.layers.31.self_attn.kv_b_proj",
+      "model.layers.31.self_attn.o_proj",
+      "model.layers.31.self_attn.q_a_layernorm",
+      "model.layers.31.self_attn.q_a_proj",
+      "model.layers.31.self_attn.q_b_proj",
+      "model.layers.32.input_layernorm",
+      "model.layers.32.mlp.gate",
+      "model.layers.32.mlp.shared_experts.down_proj",
+      "model.layers.32.mlp.shared_experts.gate_proj",
+      "model.layers.32.mlp.shared_experts.up_proj",
+      "model.layers.32.post_attention_layernorm",
+      "model.layers.32.self_attn.kv_a_layernorm",
+      "model.layers.32.self_attn.kv_a_proj_with_mqa",
+      "model.layers.32.self_attn.kv_b_proj",
+      "model.layers.32.self_attn.o_proj",
+      "model.layers.32.self_attn.q_a_layernorm",
+      "model.layers.32.self_attn.q_a_proj",
+      "model.layers.32.self_attn.q_b_proj",
+      "model.layers.33.input_layernorm",
+      "model.layers.33.mlp.gate",
+      "model.layers.33.mlp.shared_experts.down_proj",
+      "model.layers.33.mlp.shared_experts.gate_proj",
+      "model.layers.33.mlp.shared_experts.up_proj",
+      "model.layers.33.post_attention_layernorm",
+      "model.layers.33.self_attn.kv_a_layernorm",
+      "model.layers.33.self_attn.kv_a_proj_with_mqa",
+      "model.layers.33.self_attn.kv_b_proj",
+      "model.layers.33.self_attn.o_proj",
+      "model.layers.33.self_attn.q_a_layernorm",
+      "model.layers.33.self_attn.q_a_proj",
+      "model.layers.33.self_attn.q_b_proj",
+      "model.layers.34.input_layernorm",
+      "model.layers.34.mlp.gate",
+      "model.layers.34.mlp.shared_experts.down_proj",
+      "model.layers.34.mlp.shared_experts.gate_proj",
+      "model.layers.34.mlp.shared_experts.up_proj",
+      "model.layers.34.post_attention_layernorm",
+      "model.layers.34.self_attn.kv_a_layernorm",
+      "model.layers.34.self_attn.kv_a_proj_with_mqa",
+      "model.layers.34.self_attn.kv_b_proj",
+      "model.layers.34.self_attn.o_proj",
+      "model.layers.34.self_attn.q_a_layernorm",
+      "model.layers.34.self_attn.q_a_proj",
+      "model.layers.34.self_attn.q_b_proj",
+      "model.layers.35.input_layernorm",
+      "model.layers.35.mlp.gate",
+      "model.layers.35.mlp.shared_experts.down_proj",
+      "model.layers.35.mlp.shared_experts.gate_proj",
+      "model.layers.35.mlp.shared_experts.up_proj",
+      "model.layers.35.post_attention_layernorm",
+      "model.layers.35.self_attn.kv_a_layernorm",
+      "model.layers.35.self_attn.kv_a_proj_with_mqa",
+      "model.layers.35.self_attn.kv_b_proj",
+      "model.layers.35.self_attn.o_proj",
+      "model.layers.35.self_attn.q_a_layernorm",
+      "model.layers.35.self_attn.q_a_proj",
+      "model.layers.35.self_attn.q_b_proj",
+      "model.layers.36.input_layernorm",
+      "model.layers.36.mlp.gate",
+      "model.layers.36.mlp.shared_experts.down_proj",
+      "model.layers.36.mlp.shared_experts.gate_proj",
+      "model.layers.36.mlp.shared_experts.up_proj",
+      "model.layers.36.post_attention_layernorm",
+      "model.layers.36.self_attn.kv_a_layernorm",
+      "model.layers.36.self_attn.kv_a_proj_with_mqa",
+      "model.layers.36.self_attn.kv_b_proj",
+      "model.layers.36.self_attn.o_proj",
+      "model.layers.36.self_attn.q_a_layernorm",
+      "model.layers.36.self_attn.q_a_proj",
+      "model.layers.36.self_attn.q_b_proj",
+      "model.layers.37.input_layernorm",
+      "model.layers.37.mlp.gate",
+      "model.layers.37.mlp.shared_experts.down_proj",
+      "model.layers.37.mlp.shared_experts.gate_proj",
+      "model.layers.37.mlp.shared_experts.up_proj",
+      "model.layers.37.post_attention_layernorm",
+      "model.layers.37.self_attn.kv_a_layernorm",
+      "model.layers.37.self_attn.kv_a_proj_with_mqa",
+      "model.layers.37.self_attn.kv_b_proj",
+      "model.layers.37.self_attn.o_proj",
+      "model.layers.37.self_attn.q_a_layernorm",
+      "model.layers.37.self_attn.q_a_proj",
+      "model.layers.37.self_attn.q_b_proj",
+      "model.layers.38.input_layernorm",
+      "model.layers.38.mlp.gate",
+      "model.layers.38.mlp.shared_experts.down_proj",
+      "model.layers.38.mlp.shared_experts.gate_proj",
+      "model.layers.38.mlp.shared_experts.up_proj",
+      "model.layers.38.post_attention_layernorm",
+      "model.layers.38.self_attn.kv_a_layernorm",
+      "model.layers.38.self_attn.kv_a_proj_with_mqa",
+      "model.layers.38.self_attn.kv_b_proj",
+      "model.layers.38.self_attn.o_proj",
+      "model.layers.38.self_attn.q_a_layernorm",
+      "model.layers.38.self_attn.q_a_proj",
+      "model.layers.38.self_attn.q_b_proj",
+      "model.layers.39.input_layernorm",
+      "model.layers.39.mlp.gate",
+      "model.layers.39.mlp.shared_experts.down_proj",
+      "model.layers.39.mlp.shared_experts.gate_proj",
+      "model.layers.39.mlp.shared_experts.up_proj",
+      "model.layers.39.post_attention_layernorm",
+      "model.layers.39.self_attn.kv_a_layernorm",
+      "model.layers.39.self_attn.kv_a_proj_with_mqa",
+      "model.layers.39.self_attn.kv_b_proj",
+      "model.layers.39.self_attn.o_proj",
+      "model.layers.39.self_attn.q_a_layernorm",
+      "model.layers.39.self_attn.q_a_proj",
+      "model.layers.39.self_attn.q_b_proj",
+      "model.layers.4.input_layernorm",
+      "model.layers.4.mlp.gate",
+      "model.layers.4.mlp.shared_experts.down_proj",
+      "model.layers.4.mlp.shared_experts.gate_proj",
+      "model.layers.4.mlp.shared_experts.up_proj",
+      "model.layers.4.post_attention_layernorm",
+      "model.layers.4.self_attn.kv_a_layernorm",
+      "model.layers.4.self_attn.kv_a_proj_with_mqa",
+      "model.layers.4.self_attn.kv_b_proj",
+      "model.layers.4.self_attn.o_proj",
+      "model.layers.4.self_attn.q_a_layernorm",
+      "model.layers.4.self_attn.q_a_proj",
+      "model.layers.4.self_attn.q_b_proj",
+      "model.layers.40.input_layernorm",
+      "model.layers.40.mlp.gate",
+      "model.layers.40.mlp.shared_experts.down_proj",
+      "model.layers.40.mlp.shared_experts.gate_proj",
+      "model.layers.40.mlp.shared_experts.up_proj",
+      "model.layers.40.post_attention_layernorm",
+      "model.layers.40.self_attn.kv_a_layernorm",
+      "model.layers.40.self_attn.kv_a_proj_with_mqa",
+      "model.layers.40.self_attn.kv_b_proj",
+      "model.layers.40.self_attn.o_proj",
+      "model.layers.40.self_attn.q_a_layernorm",
+      "model.layers.40.self_attn.q_a_proj",
+      "model.layers.40.self_attn.q_b_proj",
+      "model.layers.41.input_layernorm",
+      "model.layers.41.mlp.gate",
+      "model.layers.41.mlp.shared_experts.down_proj",
+      "model.layers.41.mlp.shared_experts.gate_proj",
+      "model.layers.41.mlp.shared_experts.up_proj",
+      "model.layers.41.post_attention_layernorm",
+      "model.layers.41.self_attn.kv_a_layernorm",
+      "model.layers.41.self_attn.kv_a_proj_with_mqa",
+      "model.layers.41.self_attn.kv_b_proj",
+      "model.layers.41.self_attn.o_proj",
+      "model.layers.41.self_attn.q_a_layernorm",
+      "model.layers.41.self_attn.q_a_proj",
+      "model.layers.41.self_attn.q_b_proj",
+      "model.layers.42.input_layernorm",
+      "model.layers.42.mlp.gate",
+      "model.layers.42.mlp.shared_experts.down_proj",
+      "model.layers.42.mlp.shared_experts.gate_proj",
+      "model.layers.42.mlp.shared_experts.up_proj",
+      "model.layers.42.post_attention_layernorm",
+      "model.layers.42.self_attn.kv_a_layernorm",
+      "model.layers.42.self_attn.kv_a_proj_with_mqa",
+      "model.layers.42.self_attn.kv_b_proj",
+      "model.layers.42.self_attn.o_proj",
+      "model.layers.42.self_attn.q_a_layernorm",
+      "model.layers.42.self_attn.q_a_proj",
+      "model.layers.42.self_attn.q_b_proj",
+      "model.layers.43.input_layernorm",
+      "model.layers.43.mlp.gate",
+      "model.layers.43.mlp.shared_experts.down_proj",
+      "model.layers.43.mlp.shared_experts.gate_proj",
+      "model.layers.43.mlp.shared_experts.up_proj",
+      "model.layers.43.post_attention_layernorm",
+      "model.layers.43.self_attn.kv_a_layernorm",
+      "model.layers.43.self_attn.kv_a_proj_with_mqa",
+      "model.layers.43.self_attn.kv_b_proj",
+      "model.layers.43.self_attn.o_proj",
+      "model.layers.43.self_attn.q_a_layernorm",
+      "model.layers.43.self_attn.q_a_proj",
+      "model.layers.43.self_attn.q_b_proj",
+      "model.layers.44.input_layernorm",
+      "model.layers.44.mlp.gate",
+      "model.layers.44.mlp.shared_experts.down_proj",
+      "model.layers.44.mlp.shared_experts.gate_proj",
+      "model.layers.44.mlp.shared_experts.up_proj",
+      "model.layers.44.post_attention_layernorm",
+      "model.layers.44.self_attn.kv_a_layernorm",
+      "model.layers.44.self_attn.kv_a_proj_with_mqa",
+      "model.layers.44.self_attn.kv_b_proj",
+      "model.layers.44.self_attn.o_proj",
+      "model.layers.44.self_attn.q_a_layernorm",
+      "model.layers.44.self_attn.q_a_proj",
+      "model.layers.44.self_attn.q_b_proj",
+      "model.layers.45.input_layernorm",
+      "model.layers.45.mlp.gate",
+      "model.layers.45.mlp.shared_experts.down_proj",
+      "model.layers.45.mlp.shared_experts.gate_proj",
+      "model.layers.45.mlp.shared_experts.up_proj",
+      "model.layers.45.post_attention_layernorm",
+      "model.layers.45.self_attn.kv_a_layernorm",
+      "model.layers.45.self_attn.kv_a_proj_with_mqa",
+      "model.layers.45.self_attn.kv_b_proj",
+      "model.layers.45.self_attn.o_proj",
+      "model.layers.45.self_attn.q_a_layernorm",
+      "model.layers.45.self_attn.q_a_proj",
+      "model.layers.45.self_attn.q_b_proj",
+      "model.layers.46.input_layernorm",
+      "model.layers.46.mlp.gate",
+      "model.layers.46.mlp.shared_experts.down_proj",
+      "model.layers.46.mlp.shared_experts.gate_proj",
+      "model.layers.46.mlp.shared_experts.up_proj",
+      "model.layers.46.post_attention_layernorm",
+      "model.layers.46.self_attn.kv_a_layernorm",
+      "model.layers.46.self_attn.kv_a_proj_with_mqa",
+      "model.layers.46.self_attn.kv_b_proj",
+      "model.layers.46.self_attn.o_proj",
+      "model.layers.46.self_attn.q_a_layernorm",
+      "model.layers.46.self_attn.q_a_proj",
+      "model.layers.46.self_attn.q_b_proj",
+      "model.layers.47.input_layernorm",
+      "model.layers.47.mlp.gate",
+      "model.layers.47.mlp.shared_experts.down_proj",
+      "model.layers.47.mlp.shared_experts.gate_proj",
+      "model.layers.47.mlp.shared_experts.up_proj",
+      "model.layers.47.post_attention_layernorm",
+      "model.layers.47.self_attn.kv_a_layernorm",
+      "model.layers.47.self_attn.kv_a_proj_with_mqa",
+      "model.layers.47.self_attn.kv_b_proj",
+      "model.layers.47.self_attn.o_proj",
+      "model.layers.47.self_attn.q_a_layernorm",
+      "model.layers.47.self_attn.q_a_proj",
+      "model.layers.47.self_attn.q_b_proj",
+      "model.layers.48.input_layernorm",
+      "model.layers.48.mlp.gate",
+      "model.layers.48.mlp.shared_experts.down_proj",
+      "model.layers.48.mlp.shared_experts.gate_proj",
+      "model.layers.48.mlp.shared_experts.up_proj",
+      "model.layers.48.post_attention_layernorm",
+      "model.layers.48.self_attn.kv_a_layernorm",
+      "model.layers.48.self_attn.kv_a_proj_with_mqa",
+      "model.layers.48.self_attn.kv_b_proj",
+      "model.layers.48.self_attn.o_proj",
+      "model.layers.48.self_attn.q_a_layernorm",
+      "model.layers.48.self_attn.q_a_proj",
+      "model.layers.48.self_attn.q_b_proj",
+      "model.layers.49.input_layernorm",
+      "model.layers.49.mlp.gate",
+      "model.layers.49.mlp.shared_experts.down_proj",
+      "model.layers.49.mlp.shared_experts.gate_proj",
+      "model.layers.49.mlp.shared_experts.up_proj",
+      "model.layers.49.post_attention_layernorm",
+      "model.layers.49.self_attn.kv_a_layernorm",
+      "model.layers.49.self_attn.kv_a_proj_with_mqa",
+      "model.layers.49.self_attn.kv_b_proj",
+      "model.layers.49.self_attn.o_proj",
+      "model.layers.49.self_attn.q_a_layernorm",
+      "model.layers.49.self_attn.q_a_proj",
+      "model.layers.49.self_attn.q_b_proj",
+      "model.layers.5.input_layernorm",
+      "model.layers.5.mlp.gate",
+      "model.layers.5.mlp.shared_experts.down_proj",
+      "model.layers.5.mlp.shared_experts.gate_proj",
+      "model.layers.5.mlp.shared_experts.up_proj",
+      "model.layers.5.post_attention_layernorm",
+      "model.layers.5.self_attn.kv_a_layernorm",
+      "model.layers.5.self_attn.kv_a_proj_with_mqa",
+      "model.layers.5.self_attn.kv_b_proj",
+      "model.layers.5.self_attn.o_proj",
+      "model.layers.5.self_attn.q_a_layernorm",
+      "model.layers.5.self_attn.q_a_proj",
+      "model.layers.5.self_attn.q_b_proj",
+      "model.layers.50.input_layernorm",
+      "model.layers.50.mlp.gate",
+      "model.layers.50.mlp.shared_experts.down_proj",
+      "model.layers.50.mlp.shared_experts.gate_proj",
+      "model.layers.50.mlp.shared_experts.up_proj",
+      "model.layers.50.post_attention_layernorm",
+      "model.layers.50.self_attn.kv_a_layernorm",
+      "model.layers.50.self_attn.kv_a_proj_with_mqa",
+      "model.layers.50.self_attn.kv_b_proj",
+      "model.layers.50.self_attn.o_proj",
+      "model.layers.50.self_attn.q_a_layernorm",
+      "model.layers.50.self_attn.q_a_proj",
+      "model.layers.50.self_attn.q_b_proj",
+      "model.layers.51.input_layernorm",
+      "model.layers.51.mlp.gate",
+      "model.layers.51.mlp.shared_experts.down_proj",
+      "model.layers.51.mlp.shared_experts.gate_proj",
+      "model.layers.51.mlp.shared_experts.up_proj",
+      "model.layers.51.post_attention_layernorm",
+      "model.layers.51.self_attn.kv_a_layernorm",
+      "model.layers.51.self_attn.kv_a_proj_with_mqa",
+      "model.layers.51.self_attn.kv_b_proj",
+      "model.layers.51.self_attn.o_proj",
+      "model.layers.51.self_attn.q_a_layernorm",
+      "model.layers.51.self_attn.q_a_proj",
+      "model.layers.51.self_attn.q_b_proj",
+      "model.layers.52.input_layernorm",
+      "model.layers.52.mlp.gate",
+      "model.layers.52.mlp.shared_experts.down_proj",
+      "model.layers.52.mlp.shared_experts.gate_proj",
+      "model.layers.52.mlp.shared_experts.up_proj",
+      "model.layers.52.post_attention_layernorm",
+      "model.layers.52.self_attn.kv_a_layernorm",
+      "model.layers.52.self_attn.kv_a_proj_with_mqa",
+      "model.layers.52.self_attn.kv_b_proj",
+      "model.layers.52.self_attn.o_proj",
+      "model.layers.52.self_attn.q_a_layernorm",
+      "model.layers.52.self_attn.q_a_proj",
+      "model.layers.52.self_attn.q_b_proj",
+      "model.layers.53.input_layernorm",
+      "model.layers.53.mlp.gate",
+      "model.layers.53.mlp.shared_experts.down_proj",
+      "model.layers.53.mlp.shared_experts.gate_proj",
+      "model.layers.53.mlp.shared_experts.up_proj",
+      "model.layers.53.post_attention_layernorm",
+      "model.layers.53.self_attn.kv_a_layernorm",
+      "model.layers.53.self_attn.kv_a_proj_with_mqa",
+      "model.layers.53.self_attn.kv_b_proj",
+      "model.layers.53.self_attn.o_proj",
+      "model.layers.53.self_attn.q_a_layernorm",
+      "model.layers.53.self_attn.q_a_proj",
+      "model.layers.53.self_attn.q_b_proj",
+      "model.layers.54.input_layernorm",
+      "model.layers.54.mlp.gate",
+      "model.layers.54.mlp.shared_experts.down_proj",
+      "model.layers.54.mlp.shared_experts.gate_proj",
+      "model.layers.54.mlp.shared_experts.up_proj",
+      "model.layers.54.post_attention_layernorm",
+      "model.layers.54.self_attn.kv_a_layernorm",
+      "model.layers.54.self_attn.kv_a_proj_with_mqa",
+      "model.layers.54.self_attn.kv_b_proj",
+      "model.layers.54.self_attn.o_proj",
+      "model.layers.54.self_attn.q_a_layernorm",
+      "model.layers.54.self_attn.q_a_proj",
+      "model.layers.54.self_attn.q_b_proj",
+      "model.layers.55.input_layernorm",
+      "model.layers.55.mlp.gate",
+      "model.layers.55.mlp.shared_experts.down_proj",
+      "model.layers.55.mlp.shared_experts.gate_proj",
+      "model.layers.55.mlp.shared_experts.up_proj",
+      "model.layers.55.post_attention_layernorm",
+      "model.layers.55.self_attn.kv_a_layernorm",
+      "model.layers.55.self_attn.kv_a_proj_with_mqa",
+      "model.layers.55.self_attn.kv_b_proj",
+      "model.layers.55.self_attn.o_proj",
+      "model.layers.55.self_attn.q_a_layernorm",
+      "model.layers.55.self_attn.q_a_proj",
+      "model.layers.55.self_attn.q_b_proj",
+      "model.layers.56.input_layernorm",
+      "model.layers.56.mlp.gate",
+      "model.layers.56.mlp.shared_experts.down_proj",
+      "model.layers.56.mlp.shared_experts.gate_proj",
+      "model.layers.56.mlp.shared_experts.up_proj",
+      "model.layers.56.post_attention_layernorm",
+      "model.layers.56.self_attn.kv_a_layernorm",
+      "model.layers.56.self_attn.kv_a_proj_with_mqa",
+      "model.layers.56.self_attn.kv_b_proj",
+      "model.layers.56.self_attn.o_proj",
+      "model.layers.56.self_attn.q_a_layernorm",
+      "model.layers.56.self_attn.q_a_proj",
+      "model.layers.56.self_attn.q_b_proj",
+      "model.layers.57.input_layernorm",
+      "model.layers.57.mlp.gate",
+      "model.layers.57.mlp.shared_experts.down_proj",
+      "model.layers.57.mlp.shared_experts.gate_proj",
+      "model.layers.57.mlp.shared_experts.up_proj",
+      "model.layers.57.post_attention_layernorm",
+      "model.layers.57.self_attn.kv_a_layernorm",
+      "model.layers.57.self_attn.kv_a_proj_with_mqa",
+      "model.layers.57.self_attn.kv_b_proj",
+      "model.layers.57.self_attn.o_proj",
+      "model.layers.57.self_attn.q_a_layernorm",
+      "model.layers.57.self_attn.q_a_proj",
+      "model.layers.57.self_attn.q_b_proj",
+      "model.layers.58.input_layernorm",
+      "model.layers.58.mlp.gate",
+      "model.layers.58.mlp.shared_experts.down_proj",
+      "model.layers.58.mlp.shared_experts.gate_proj",
+      "model.layers.58.mlp.shared_experts.up_proj",
+      "model.layers.58.post_attention_layernorm",
+      "model.layers.58.self_attn.kv_a_layernorm",
+      "model.layers.58.self_attn.kv_a_proj_with_mqa",
+      "model.layers.58.self_attn.kv_b_proj",
+      "model.layers.58.self_attn.o_proj",
+      "model.layers.58.self_attn.q_a_layernorm",
+      "model.layers.58.self_attn.q_a_proj",
+      "model.layers.58.self_attn.q_b_proj",
+      "model.layers.59.input_layernorm",
+      "model.layers.59.mlp.gate",
+      "model.layers.59.mlp.shared_experts.down_proj",
+      "model.layers.59.mlp.shared_experts.gate_proj",
+      "model.layers.59.mlp.shared_experts.up_proj",
+      "model.layers.59.post_attention_layernorm",
+      "model.layers.59.self_attn.kv_a_layernorm",
+      "model.layers.59.self_attn.kv_a_proj_with_mqa",
+      "model.layers.59.self_attn.kv_b_proj",
+      "model.layers.59.self_attn.o_proj",
+      "model.layers.59.self_attn.q_a_layernorm",
+      "model.layers.59.self_attn.q_a_proj",
+      "model.layers.59.self_attn.q_b_proj",
+      "model.layers.6.input_layernorm",
+      "model.layers.6.mlp.gate",
+      "model.layers.6.mlp.shared_experts.down_proj",
+      "model.layers.6.mlp.shared_experts.gate_proj",
+      "model.layers.6.mlp.shared_experts.up_proj",
+      "model.layers.6.post_attention_layernorm",
+      "model.layers.6.self_attn.kv_a_layernorm",
+      "model.layers.6.self_attn.kv_a_proj_with_mqa",
+      "model.layers.6.self_attn.kv_b_proj",
+      "model.layers.6.self_attn.o_proj",
+      "model.layers.6.self_attn.q_a_layernorm",
+      "model.layers.6.self_attn.q_a_proj",
+      "model.layers.6.self_attn.q_b_proj",
+      "model.layers.60.input_layernorm",
+      "model.layers.60.mlp.gate",
+      "model.layers.60.mlp.shared_experts.down_proj",
+      "model.layers.60.mlp.shared_experts.gate_proj",
+      "model.layers.60.mlp.shared_experts.up_proj",
+      "model.layers.60.post_attention_layernorm",
+      "model.layers.60.self_attn.kv_a_layernorm",
+      "model.layers.60.self_attn.kv_a_proj_with_mqa",
+      "model.layers.60.self_attn.kv_b_proj",
+      "model.layers.60.self_attn.o_proj",
+      "model.layers.60.self_attn.q_a_layernorm",
+      "model.layers.60.self_attn.q_a_proj",
+      "model.layers.60.self_attn.q_b_proj",
+      "model.layers.61.input_layernorm",
+      "model.layers.61.mlp.gate",
+      "model.layers.61.mlp.shared_experts.down_proj",
+      "model.layers.61.mlp.shared_experts.gate_proj",
+      "model.layers.61.mlp.shared_experts.up_proj",
+      "model.layers.61.post_attention_layernorm",
+      "model.layers.61.self_attn.kv_a_layernorm",
+      "model.layers.61.self_attn.kv_a_proj_with_mqa",
+      "model.layers.61.self_attn.kv_b_proj",
+      "model.layers.61.self_attn.o_proj",
+      "model.layers.61.self_attn.q_a_layernorm",
+      "model.layers.61.self_attn.q_a_proj",
+      "model.layers.61.self_attn.q_b_proj",
+      "model.layers.62.input_layernorm",
+      "model.layers.62.mlp.gate",
+      "model.layers.62.mlp.shared_experts.down_proj",
+      "model.layers.62.mlp.shared_experts.gate_proj",
+      "model.layers.62.mlp.shared_experts.up_proj",
+      "model.layers.62.post_attention_layernorm",
+      "model.layers.62.self_attn.kv_a_layernorm",
+      "model.layers.62.self_attn.kv_a_proj_with_mqa",
+      "model.layers.62.self_attn.kv_b_proj",
+      "model.layers.62.self_attn.o_proj",
+      "model.layers.62.self_attn.q_a_layernorm",
+      "model.layers.62.self_attn.q_a_proj",
+      "model.layers.62.self_attn.q_b_proj",
+      "model.layers.63.input_layernorm",
+      "model.layers.63.mlp.gate",
+      "model.layers.63.mlp.shared_experts.down_proj",
+      "model.layers.63.mlp.shared_experts.gate_proj",
+      "model.layers.63.mlp.shared_experts.up_proj",
+      "model.layers.63.post_attention_layernorm",
+      "model.layers.63.self_attn.kv_a_layernorm",
+      "model.layers.63.self_attn.kv_a_proj_with_mqa",
+      "model.layers.63.self_attn.kv_b_proj",
+      "model.layers.63.self_attn.o_proj",
+      "model.layers.63.self_attn.q_a_layernorm",
+      "model.layers.63.self_attn.q_a_proj",
+      "model.layers.63.self_attn.q_b_proj",
+      "model.layers.64.input_layernorm",
+      "model.layers.64.mlp.gate",
+      "model.layers.64.mlp.shared_experts.down_proj",
+      "model.layers.64.mlp.shared_experts.gate_proj",
+      "model.layers.64.mlp.shared_experts.up_proj",
+      "model.layers.64.post_attention_layernorm",
+      "model.layers.64.self_attn.kv_a_layernorm",
+      "model.layers.64.self_attn.kv_a_proj_with_mqa",
+      "model.layers.64.self_attn.kv_b_proj",
+      "model.layers.64.self_attn.o_proj",
+      "model.layers.64.self_attn.q_a_layernorm",
+      "model.layers.64.self_attn.q_a_proj",
+      "model.layers.64.self_attn.q_b_proj",
+      "model.layers.65.input_layernorm",
+      "model.layers.65.mlp.gate",
+      "model.layers.65.mlp.shared_experts.down_proj",
+      "model.layers.65.mlp.shared_experts.gate_proj",
+      "model.layers.65.mlp.shared_experts.up_proj",
+      "model.layers.65.post_attention_layernorm",
+      "model.layers.65.self_attn.kv_a_layernorm",
+      "model.layers.65.self_attn.kv_a_proj_with_mqa",
+      "model.layers.65.self_attn.kv_b_proj",
+      "model.layers.65.self_attn.o_proj",
+      "model.layers.65.self_attn.q_a_layernorm",
+      "model.layers.65.self_attn.q_a_proj",
+      "model.layers.65.self_attn.q_b_proj",
+      "model.layers.66.input_layernorm",
+      "model.layers.66.mlp.gate",
+      "model.layers.66.mlp.shared_experts.down_proj",
+      "model.layers.66.mlp.shared_experts.gate_proj",
+      "model.layers.66.mlp.shared_experts.up_proj",
+      "model.layers.66.post_attention_layernorm",
+      "model.layers.66.self_attn.kv_a_layernorm",
+      "model.layers.66.self_attn.kv_a_proj_with_mqa",
+      "model.layers.66.self_attn.kv_b_proj",
+      "model.layers.66.self_attn.o_proj",
+      "model.layers.66.self_attn.q_a_layernorm",
+      "model.layers.66.self_attn.q_a_proj",
+      "model.layers.66.self_attn.q_b_proj",
+      "model.layers.67.input_layernorm",
+      "model.layers.67.mlp.gate",
+      "model.layers.67.mlp.shared_experts.down_proj",
+      "model.layers.67.mlp.shared_experts.gate_proj",
+      "model.layers.67.mlp.shared_experts.up_proj",
+      "model.layers.67.post_attention_layernorm",
+      "model.layers.67.self_attn.kv_a_layernorm",
+      "model.layers.67.self_attn.kv_a_proj_with_mqa",
+      "model.layers.67.self_attn.kv_b_proj",
+      "model.layers.67.self_attn.o_proj",
+      "model.layers.67.self_attn.q_a_layernorm",
+      "model.layers.67.self_attn.q_a_proj",
+      "model.layers.67.self_attn.q_b_proj",
+      "model.layers.68.input_layernorm",
+      "model.layers.68.mlp.gate",
+      "model.layers.68.mlp.shared_experts.down_proj",
+      "model.layers.68.mlp.shared_experts.gate_proj",
+      "model.layers.68.mlp.shared_experts.up_proj",
+      "model.layers.68.post_attention_layernorm",
+      "model.layers.68.self_attn.kv_a_layernorm",
+      "model.layers.68.self_attn.kv_a_proj_with_mqa",
+      "model.layers.68.self_attn.kv_b_proj",
+      "model.layers.68.self_attn.o_proj",
+      "model.layers.68.self_attn.q_a_layernorm",
+      "model.layers.68.self_attn.q_a_proj",
+      "model.layers.68.self_attn.q_b_proj",
+      "model.layers.69.input_layernorm",
+      "model.layers.69.mlp.gate",
+      "model.layers.69.mlp.shared_experts.down_proj",
+      "model.layers.69.mlp.shared_experts.gate_proj",
+      "model.layers.69.mlp.shared_experts.up_proj",
+      "model.layers.69.post_attention_layernorm",
+      "model.layers.69.self_attn.kv_a_layernorm",
+      "model.layers.69.self_attn.kv_a_proj_with_mqa",
+      "model.layers.69.self_attn.kv_b_proj",
+      "model.layers.69.self_attn.o_proj",
+      "model.layers.69.self_attn.q_a_layernorm",
+      "model.layers.69.self_attn.q_a_proj",
+      "model.layers.69.self_attn.q_b_proj",
+      "model.layers.7.input_layernorm",
+      "model.layers.7.mlp.gate",
+      "model.layers.7.mlp.shared_experts.down_proj",
+      "model.layers.7.mlp.shared_experts.gate_proj",
+      "model.layers.7.mlp.shared_experts.up_proj",
+      "model.layers.7.post_attention_layernorm",
+      "model.layers.7.self_attn.kv_a_layernorm",
+      "model.layers.7.self_attn.kv_a_proj_with_mqa",
+      "model.layers.7.self_attn.kv_b_proj",
+      "model.layers.7.self_attn.o_proj",
+      "model.layers.7.self_attn.q_a_layernorm",
+      "model.layers.7.self_attn.q_a_proj",
+      "model.layers.7.self_attn.q_b_proj",
+      "model.layers.70.input_layernorm",
+      "model.layers.70.mlp.gate",
+      "model.layers.70.mlp.shared_experts.down_proj",
+      "model.layers.70.mlp.shared_experts.gate_proj",
+      "model.layers.70.mlp.shared_experts.up_proj",
+      "model.layers.70.post_attention_layernorm",
+      "model.layers.70.self_attn.kv_a_layernorm",
+      "model.layers.70.self_attn.kv_a_proj_with_mqa",
+      "model.layers.70.self_attn.kv_b_proj",
+      "model.layers.70.self_attn.o_proj",
+      "model.layers.70.self_attn.q_a_layernorm",
+      "model.layers.70.self_attn.q_a_proj",
+      "model.layers.70.self_attn.q_b_proj",
+      "model.layers.71.input_layernorm",
+      "model.layers.71.mlp.gate",
+      "model.layers.71.mlp.shared_experts.down_proj",
+      "model.layers.71.mlp.shared_experts.gate_proj",
+      "model.layers.71.mlp.shared_experts.up_proj",
+      "model.layers.71.post_attention_layernorm",
+      "model.layers.71.self_attn.kv_a_layernorm",
+      "model.layers.71.self_attn.kv_a_proj_with_mqa",
+      "model.layers.71.self_attn.kv_b_proj",
+      "model.layers.71.self_attn.o_proj",
+      "model.layers.71.self_attn.q_a_layernorm",
+      "model.layers.71.self_attn.q_a_proj",
+      "model.layers.71.self_attn.q_b_proj",
+      "model.layers.72.input_layernorm",
+      "model.layers.72.mlp.gate",
+      "model.layers.72.mlp.shared_experts.down_proj",
+      "model.layers.72.mlp.shared_experts.gate_proj",
+      "model.layers.72.mlp.shared_experts.up_proj",
+      "model.layers.72.post_attention_layernorm",
+      "model.layers.72.self_attn.kv_a_layernorm",
+      "model.layers.72.self_attn.kv_a_proj_with_mqa",
+      "model.layers.72.self_attn.kv_b_proj",
+      "model.layers.72.self_attn.o_proj",
+      "model.layers.72.self_attn.q_a_layernorm",
+      "model.layers.72.self_attn.q_a_proj",
+      "model.layers.72.self_attn.q_b_proj",
+      "model.layers.73.input_layernorm",
+      "model.layers.73.mlp.gate",
+      "model.layers.73.mlp.shared_experts.down_proj",
+      "model.layers.73.mlp.shared_experts.gate_proj",
+      "model.layers.73.mlp.shared_experts.up_proj",
+      "model.layers.73.post_attention_layernorm",
+      "model.layers.73.self_attn.kv_a_layernorm",
+      "model.layers.73.self_attn.kv_a_proj_with_mqa",
+      "model.layers.73.self_attn.kv_b_proj",
+      "model.layers.73.self_attn.o_proj",
+      "model.layers.73.self_attn.q_a_layernorm",
+      "model.layers.73.self_attn.q_a_proj",
+      "model.layers.73.self_attn.q_b_proj",
+      "model.layers.74.input_layernorm",
+      "model.layers.74.mlp.gate",
+      "model.layers.74.mlp.shared_experts.down_proj",
+      "model.layers.74.mlp.shared_experts.gate_proj",
+      "model.layers.74.mlp.shared_experts.up_proj",
+      "model.layers.74.post_attention_layernorm",
+      "model.layers.74.self_attn.kv_a_layernorm",
+      "model.layers.74.self_attn.kv_a_proj_with_mqa",
+      "model.layers.74.self_attn.kv_b_proj",
+      "model.layers.74.self_attn.o_proj",
+      "model.layers.74.self_attn.q_a_layernorm",
+      "model.layers.74.self_attn.q_a_proj",
+      "model.layers.74.self_attn.q_b_proj",
+      "model.layers.75.input_layernorm",
+      "model.layers.75.mlp.gate",
+      "model.layers.75.mlp.shared_experts.down_proj",
+      "model.layers.75.mlp.shared_experts.gate_proj",
+      "model.layers.75.mlp.shared_experts.up_proj",
+      "model.layers.75.post_attention_layernorm",
+      "model.layers.75.self_attn.kv_a_layernorm",
+      "model.layers.75.self_attn.kv_a_proj_with_mqa",
+      "model.layers.75.self_attn.kv_b_proj",
+      "model.layers.75.self_attn.o_proj",
+      "model.layers.75.self_attn.q_a_layernorm",
+      "model.layers.75.self_attn.q_a_proj",
+      "model.layers.75.self_attn.q_b_proj",
+      "model.layers.76.input_layernorm",
+      "model.layers.76.mlp.gate",
+      "model.layers.76.mlp.shared_experts.down_proj",
+      "model.layers.76.mlp.shared_experts.gate_proj",
+      "model.layers.76.mlp.shared_experts.up_proj",
+      "model.layers.76.post_attention_layernorm",
+      "model.layers.76.self_attn.kv_a_layernorm",
+      "model.layers.76.self_attn.kv_a_proj_with_mqa",
+      "model.layers.76.self_attn.kv_b_proj",
+      "model.layers.76.self_attn.o_proj",
+      "model.layers.76.self_attn.q_a_layernorm",
+      "model.layers.76.self_attn.q_a_proj",
+      "model.layers.76.self_attn.q_b_proj",
+      "model.layers.77.input_layernorm",
+      "model.layers.77.mlp.gate",
+      "model.layers.77.mlp.shared_experts.down_proj",
+      "model.layers.77.mlp.shared_experts.gate_proj",
+      "model.layers.77.mlp.shared_experts.up_proj",
+      "model.layers.77.post_attention_layernorm",
+      "model.layers.77.self_attn.kv_a_layernorm",
+      "model.layers.77.self_attn.kv_a_proj_with_mqa",
+      "model.layers.77.self_attn.kv_b_proj",
+      "model.layers.77.self_attn.o_proj",
+      "model.layers.77.self_attn.q_a_layernorm",
+      "model.layers.77.self_attn.q_a_proj",
+      "model.layers.77.self_attn.q_b_proj",
+      "model.layers.78.input_layernorm",
+      "model.layers.78.mlp.gate",
+      "model.layers.78.mlp.shared_experts.down_proj",
+      "model.layers.78.mlp.shared_experts.gate_proj",
+      "model.layers.78.mlp.shared_experts.up_proj",
+      "model.layers.78.post_attention_layernorm",
+      "model.layers.78.self_attn.kv_a_layernorm",
+      "model.layers.78.self_attn.kv_a_proj_with_mqa",
+      "model.layers.78.self_attn.kv_b_proj",
+      "model.layers.78.self_attn.o_proj",
+      "model.layers.78.self_attn.q_a_layernorm",
+      "model.layers.78.self_attn.q_a_proj",
+      "model.layers.78.self_attn.q_b_proj",
+      "model.layers.79.input_layernorm",
+      "model.layers.79.mlp.gate",
+      "model.layers.79.mlp.shared_experts.down_proj",
+      "model.layers.79.mlp.shared_experts.gate_proj",
+      "model.layers.79.mlp.shared_experts.up_proj",
+      "model.layers.79.post_attention_layernorm",
+      "model.layers.79.self_attn.kv_a_layernorm",
+      "model.layers.79.self_attn.kv_a_proj_with_mqa",
+      "model.layers.79.self_attn.kv_b_proj",
+      "model.layers.79.self_attn.o_proj",
+      "model.layers.79.self_attn.q_a_layernorm",
+      "model.layers.79.self_attn.q_a_proj",
+      "model.layers.79.self_attn.q_b_proj",
+      "model.layers.8.input_layernorm",
+      "model.layers.8.mlp.gate",
+      "model.layers.8.mlp.shared_experts.down_proj",
+      "model.layers.8.mlp.shared_experts.gate_proj",
+      "model.layers.8.mlp.shared_experts.up_proj",
+      "model.layers.8.post_attention_layernorm",
+      "model.layers.8.self_attn.kv_a_layernorm",
+      "model.layers.8.self_attn.kv_a_proj_with_mqa",
+      "model.layers.8.self_attn.kv_b_proj",
+      "model.layers.8.self_attn.o_proj",
+      "model.layers.8.self_attn.q_a_layernorm",
+      "model.layers.8.self_attn.q_a_proj",
+      "model.layers.8.self_attn.q_b_proj",
+      "model.layers.80.input_layernorm",
+      "model.layers.80.mlp.gate",
+      "model.layers.80.mlp.shared_experts.down_proj",
+      "model.layers.80.mlp.shared_experts.gate_proj",
+      "model.layers.80.mlp.shared_experts.up_proj",
+      "model.layers.80.post_attention_layernorm",
+      "model.layers.80.self_attn.kv_a_layernorm",
+      "model.layers.80.self_attn.kv_a_proj_with_mqa",
+      "model.layers.80.self_attn.kv_b_proj",
+      "model.layers.80.self_attn.o_proj",
+      "model.layers.80.self_attn.q_a_layernorm",
+      "model.layers.80.self_attn.q_a_proj",
+      "model.layers.80.self_attn.q_b_proj",
+      "model.layers.81.input_layernorm",
+      "model.layers.81.mlp.gate",
+      "model.layers.81.mlp.shared_experts.down_proj",
+      "model.layers.81.mlp.shared_experts.gate_proj",
+      "model.layers.81.mlp.shared_experts.up_proj",
+      "model.layers.81.post_attention_layernorm",
+      "model.layers.81.self_attn.kv_a_layernorm",
+      "model.layers.81.self_attn.kv_a_proj_with_mqa",
+      "model.layers.81.self_attn.kv_b_proj",
+      "model.layers.81.self_attn.o_proj",
+      "model.layers.81.self_attn.q_a_layernorm",
+      "model.layers.81.self_attn.q_a_proj",
+      "model.layers.81.self_attn.q_b_proj",
+      "model.layers.82.input_layernorm",
+      "model.layers.82.mlp.gate",
+      "model.layers.82.mlp.shared_experts.down_proj",
+      "model.layers.82.mlp.shared_experts.gate_proj",
+      "model.layers.82.mlp.shared_experts.up_proj",
+      "model.layers.82.post_attention_layernorm",
+      "model.layers.82.self_attn.kv_a_layernorm",
+      "model.layers.82.self_attn.kv_a_proj_with_mqa",
+      "model.layers.82.self_attn.kv_b_proj",
+      "model.layers.82.self_attn.o_proj",
+      "model.layers.82.self_attn.q_a_layernorm",
+      "model.layers.82.self_attn.q_a_proj",
+      "model.layers.82.self_attn.q_b_proj",
+      "model.layers.83.input_layernorm",
+      "model.layers.83.mlp.gate",
+      "model.layers.83.mlp.shared_experts.down_proj",
+      "model.layers.83.mlp.shared_experts.gate_proj",
+      "model.layers.83.mlp.shared_experts.up_proj",
+      "model.layers.83.post_attention_layernorm",
+      "model.layers.83.self_attn.kv_a_layernorm",
+      "model.layers.83.self_attn.kv_a_proj_with_mqa",
+      "model.layers.83.self_attn.kv_b_proj",
+      "model.layers.83.self_attn.o_proj",
+      "model.layers.83.self_attn.q_a_layernorm",
+      "model.layers.83.self_attn.q_a_proj",
+      "model.layers.83.self_attn.q_b_proj",
+      "model.layers.84.input_layernorm",
+      "model.layers.84.mlp.gate",
+      "model.layers.84.mlp.shared_experts.down_proj",
+      "model.layers.84.mlp.shared_experts.gate_proj",
+      "model.layers.84.mlp.shared_experts.up_proj",
+      "model.layers.84.post_attention_layernorm",
+      "model.layers.84.self_attn.kv_a_layernorm",
+      "model.layers.84.self_attn.kv_a_proj_with_mqa",
+      "model.layers.84.self_attn.kv_b_proj",
+      "model.layers.84.self_attn.o_proj",
+      "model.layers.84.self_attn.q_a_layernorm",
+      "model.layers.84.self_attn.q_a_proj",
+      "model.layers.84.self_attn.q_b_proj",
+      "model.layers.85.input_layernorm",
+      "model.layers.85.mlp.gate",
+      "model.layers.85.mlp.shared_experts.down_proj",
+      "model.layers.85.mlp.shared_experts.gate_proj",
+      "model.layers.85.mlp.shared_experts.up_proj",
+      "model.layers.85.post_attention_layernorm",
+      "model.layers.85.self_attn.kv_a_layernorm",
+      "model.layers.85.self_attn.kv_a_proj_with_mqa",
+      "model.layers.85.self_attn.kv_b_proj",
+      "model.layers.85.self_attn.o_proj",
+      "model.layers.85.self_attn.q_a_layernorm",
+      "model.layers.85.self_attn.q_a_proj",
+      "model.layers.85.self_attn.q_b_proj",
+      "model.layers.86.input_layernorm",
+      "model.layers.86.mlp.gate",
+      "model.layers.86.mlp.shared_experts.down_proj",
+      "model.layers.86.mlp.shared_experts.gate_proj",
+      "model.layers.86.mlp.shared_experts.up_proj",
+      "model.layers.86.post_attention_layernorm",
+      "model.layers.86.self_attn.kv_a_layernorm",
+      "model.layers.86.self_attn.kv_a_proj_with_mqa",
+      "model.layers.86.self_attn.kv_b_proj",
+      "model.layers.86.self_attn.o_proj",
+      "model.layers.86.self_attn.q_a_layernorm",
+      "model.layers.86.self_attn.q_a_proj",
+      "model.layers.86.self_attn.q_b_proj",
+      "model.layers.87.input_layernorm",
+      "model.layers.87.mlp.gate",
+      "model.layers.87.mlp.shared_experts.down_proj",
+      "model.layers.87.mlp.shared_experts.gate_proj",
+      "model.layers.87.mlp.shared_experts.up_proj",
+      "model.layers.87.post_attention_layernorm",
+      "model.layers.87.self_attn.kv_a_layernorm",
+      "model.layers.87.self_attn.kv_a_proj_with_mqa",
+      "model.layers.87.self_attn.kv_b_proj",
+      "model.layers.87.self_attn.o_proj",
+      "model.layers.87.self_attn.q_a_layernorm",
+      "model.layers.87.self_attn.q_a_proj",
+      "model.layers.87.self_attn.q_b_proj",
+      "model.layers.88.input_layernorm",
+      "model.layers.88.mlp.gate",
+      "model.layers.88.mlp.shared_experts.down_proj",
+      "model.layers.88.mlp.shared_experts.gate_proj",
+      "model.layers.88.mlp.shared_experts.up_proj",
+      "model.layers.88.post_attention_layernorm",
+      "model.layers.88.self_attn.kv_a_layernorm",
+      "model.layers.88.self_attn.kv_a_proj_with_mqa",
+      "model.layers.88.self_attn.kv_b_proj",
+      "model.layers.88.self_attn.o_proj",
+      "model.layers.88.self_attn.q_a_layernorm",
+      "model.layers.88.self_attn.q_a_proj",
+      "model.layers.88.self_attn.q_b_proj",
+      "model.layers.89.input_layernorm",
+      "model.layers.89.mlp.gate",
+      "model.layers.89.mlp.shared_experts.down_proj",
+      "model.layers.89.mlp.shared_experts.gate_proj",
+      "model.layers.89.mlp.shared_experts.up_proj",
+      "model.layers.89.post_attention_layernorm",
+      "model.layers.89.self_attn.kv_a_layernorm",
+      "model.layers.89.self_attn.kv_a_proj_with_mqa",
+      "model.layers.89.self_attn.kv_b_proj",
+      "model.layers.89.self_attn.o_proj",
+      "model.layers.89.self_attn.q_a_layernorm",
+      "model.layers.89.self_attn.q_a_proj",
+      "model.layers.89.self_attn.q_b_proj",
+      "model.layers.9.input_layernorm",
+      "model.layers.9.mlp.gate",
+      "model.layers.9.mlp.shared_experts.down_proj",
+      "model.layers.9.mlp.shared_experts.gate_proj",
+      "model.layers.9.mlp.shared_experts.up_proj",
+      "model.layers.9.post_attention_layernorm",
+      "model.layers.9.self_attn.kv_a_layernorm",
+      "model.layers.9.self_attn.kv_a_proj_with_mqa",
+      "model.layers.9.self_attn.kv_b_proj",
+      "model.layers.9.self_attn.o_proj",
+      "model.layers.9.self_attn.q_a_layernorm",
+      "model.layers.9.self_attn.q_a_proj",
+      "model.layers.9.self_attn.q_b_proj",
+      "model.layers.90.input_layernorm",
+      "model.layers.90.mlp.gate",
+      "model.layers.90.mlp.shared_experts.down_proj",
+      "model.layers.90.mlp.shared_experts.gate_proj",
+      "model.layers.90.mlp.shared_experts.up_proj",
+      "model.layers.90.post_attention_layernorm",
+      "model.layers.90.self_attn.kv_a_layernorm",
+      "model.layers.90.self_attn.kv_a_proj_with_mqa",
+      "model.layers.90.self_attn.kv_b_proj",
+      "model.layers.90.self_attn.o_proj",
+      "model.layers.90.self_attn.q_a_layernorm",
+      "model.layers.90.self_attn.q_a_proj",
+      "model.layers.90.self_attn.q_b_proj",
+      "model.layers.91.input_layernorm",
+      "model.layers.91.mlp.gate",
+      "model.layers.91.mlp.shared_experts.down_proj",
+      "model.layers.91.mlp.shared_experts.gate_proj",
+      "model.layers.91.mlp.shared_experts.up_proj",
+      "model.layers.91.post_attention_layernorm",
+      "model.layers.91.self_attn.kv_a_layernorm",
+      "model.layers.91.self_attn.kv_a_proj_with_mqa",
+      "model.layers.91.self_attn.kv_b_proj",
+      "model.layers.91.self_attn.o_proj",
+      "model.layers.91.self_attn.q_a_layernorm",
+      "model.layers.91.self_attn.q_a_proj",
+      "model.layers.91.self_attn.q_b_proj",
+      "model.layers.92.input_layernorm",
+      "model.layers.92.mlp.gate",
+      "model.layers.92.mlp.shared_experts.down_proj",
+      "model.layers.92.mlp.shared_experts.gate_proj",
+      "model.layers.92.mlp.shared_experts.up_proj",
+      "model.layers.92.post_attention_layernorm",
+      "model.layers.92.self_attn.kv_a_layernorm",
+      "model.layers.92.self_attn.kv_a_proj_with_mqa",
+      "model.layers.92.self_attn.kv_b_proj",
+      "model.layers.92.self_attn.o_proj",
+      "model.layers.92.self_attn.q_a_layernorm",
+      "model.layers.92.self_attn.q_a_proj",
+      "model.layers.92.self_attn.q_b_proj",
+      "model.layers.93.input_layernorm",
+      "model.layers.93.mlp.gate",
+      "model.layers.93.mlp.shared_experts.down_proj",
+      "model.layers.93.mlp.shared_experts.gate_proj",
+      "model.layers.93.mlp.shared_experts.up_proj",
+      "model.layers.93.post_attention_layernorm",
+      "model.layers.93.self_attn.kv_a_layernorm",
+      "model.layers.93.self_attn.kv_a_proj_with_mqa",
+      "model.layers.93.self_attn.kv_b_proj",
+      "model.layers.93.self_attn.o_proj",
+      "model.layers.93.self_attn.q_a_layernorm",
+      "model.layers.93.self_attn.q_a_proj",
+      "model.layers.93.self_attn.q_b_proj",
+      "model.layers.94.input_layernorm",
+      "model.layers.94.mlp.gate",
+      "model.layers.94.mlp.shared_experts.down_proj",
+      "model.layers.94.mlp.shared_experts.gate_proj",
+      "model.layers.94.mlp.shared_experts.up_proj",
+      "model.layers.94.post_attention_layernorm",
+      "model.layers.94.self_attn.kv_a_layernorm",
+      "model.layers.94.self_attn.kv_a_proj_with_mqa",
+      "model.layers.94.self_attn.kv_b_proj",
+      "model.layers.94.self_attn.o_proj",
+      "model.layers.94.self_attn.q_a_layernorm",
+      "model.layers.94.self_attn.q_a_proj",
+      "model.layers.94.self_attn.q_b_proj",
+      "model.layers.95.input_layernorm",
+      "model.layers.95.mlp.gate",
+      "model.layers.95.mlp.shared_experts.down_proj",
+      "model.layers.95.mlp.shared_experts.gate_proj",
+      "model.layers.95.mlp.shared_experts.up_proj",
+      "model.layers.95.post_attention_layernorm",
+      "model.layers.95.self_attn.kv_a_layernorm",
+      "model.layers.95.self_attn.kv_a_proj_with_mqa",
+      "model.layers.95.self_attn.kv_b_proj",
+      "model.layers.95.self_attn.o_proj",
+      "model.layers.95.self_attn.q_a_layernorm",
+      "model.layers.95.self_attn.q_a_proj",
+      "model.layers.95.self_attn.q_b_proj",
+      "model.layers.96.input_layernorm",
+      "model.layers.96.mlp.gate",
+      "model.layers.96.mlp.shared_experts.down_proj",
+      "model.layers.96.mlp.shared_experts.gate_proj",
+      "model.layers.96.mlp.shared_experts.up_proj",
+      "model.layers.96.post_attention_layernorm",
+      "model.layers.96.self_attn.kv_a_layernorm",
+      "model.layers.96.self_attn.kv_a_proj_with_mqa",
+      "model.layers.96.self_attn.kv_b_proj",
+      "model.layers.96.self_attn.o_proj",
+      "model.layers.96.self_attn.q_a_layernorm",
+      "model.layers.96.self_attn.q_a_proj",
+      "model.layers.96.self_attn.q_b_proj",
+      "model.layers.97.input_layernorm",
+      "model.layers.97.mlp.gate",
+      "model.layers.97.mlp.shared_experts.down_proj",
+      "model.layers.97.mlp.shared_experts.gate_proj",
+      "model.layers.97.mlp.shared_experts.up_proj",
+      "model.layers.97.post_attention_layernorm",
+      "model.layers.97.self_attn.kv_a_layernorm",
+      "model.layers.97.self_attn.kv_a_proj_with_mqa",
+      "model.layers.97.self_attn.kv_b_proj",
+      "model.layers.97.self_attn.o_proj",
+      "model.layers.97.self_attn.q_a_layernorm",
+      "model.layers.97.self_attn.q_a_proj",
+      "model.layers.97.self_attn.q_b_proj",
+      "model.layers.98.input_layernorm",
+      "model.layers.98.mlp.gate",
+      "model.layers.98.mlp.shared_experts.down_proj",
+      "model.layers.98.mlp.shared_experts.gate_proj",
+      "model.layers.98.mlp.shared_experts.up_proj",
+      "model.layers.98.post_attention_layernorm",
+      "model.layers.98.self_attn.kv_a_layernorm",
+      "model.layers.98.self_attn.kv_a_proj_with_mqa",
+      "model.layers.98.self_attn.kv_b_proj",
+      "model.layers.98.self_attn.o_proj",
+      "model.layers.98.self_attn.q_a_layernorm",
+      "model.layers.98.self_attn.q_a_proj",
+      "model.layers.98.self_attn.q_b_proj",
+      "model.layers.99.input_layernorm",
+      "model.layers.99.mlp.gate",
+      "model.layers.99.mlp.shared_experts.down_proj",
+      "model.layers.99.mlp.shared_experts.gate_proj",
+      "model.layers.99.mlp.shared_experts.up_proj",
+      "model.layers.99.post_attention_layernorm",
+      "model.layers.99.self_attn.kv_a_layernorm",
+      "model.layers.99.self_attn.kv_a_proj_with_mqa",
+      "model.layers.99.self_attn.kv_b_proj",
+      "model.layers.99.self_attn.o_proj",
+      "model.layers.99.self_attn.q_a_layernorm",
+      "model.layers.99.self_attn.q_a_proj",
+      "model.layers.99.self_attn.q_b_proj",
+      "model.norm"
+    ],
+    "quant_method": "fp8",
+    "weight_block_size": [
+      128,
+      128
+    ]
+  },
+  "remove_invalid_values": false,
+  "repetition_penalty": 1.0,
+  "return_dict": true,
+  "return_dict_in_generate": false,
+  "rms_norm_eps": 1e-06,
+  "rope_interleave": true,
+  "rope_scaling": {
+    "beta_fast": 1.0,
+    "beta_slow": 1.0,
+    "factor": 32.0,
+    "mscale": 1.0,
+    "mscale_all_dim": 1.0,
+    "original_max_position_embeddings": 4096,
+    "rope_type": "yarn",
+    "type": "yarn"
+  },
+  "rope_theta": 50000.0,
+  "routed_scaling_factor": 2.827,
+  "scoring_func": "sigmoid",
+  "sep_token_id": null,
+  "seq_aux": true,
+  "suppress_tokens": null,
+  "task_specific_params": null,
+  "temperature": 1.0,
+  "tf_legacy_loss": false,
+  "tie_encoder_decoder": false,
+  "tie_word_embeddings": false,
+  "tokenizer_class": null,
+  "top_k": 50,
+  "top_p": 1.0,
+  "topk_group": 1,
+  "topk_method": "noaux_tc",
+  "torch_dtype": "bfloat16",
+  "torchscript": false,
+  "transformers_version": "4.55.4",
+  "typical_p": 1.0,
+  "use_bfloat16": false,
+  "use_cache": true,
+  "v_head_dim": 128,
+  "vocab_size": 163840
+}

configuration_deepseek.py

@@ -0,0 +1,212 @@
+# Copy from https://huggingface.co/deepseek-ai/DeepSeek-V3/blob/main/configuration_deepseek.py
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+DEEPSEEK_PRETRAINED_CONFIG_ARCHIVE_MAP = {}
+class DeepseekV3Config(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`DeepseekV3Model`]. It is used to instantiate an DeepSeek
+    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
+    defaults will yield a similar configuration to that of the DeepSeek-V3.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 129280):
+            Vocabulary size of the Deep model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`DeepseekV3Model`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 11008):
+            Dimension of the MLP representations.
+        moe_intermediate_size (`int`, *optional*, defaults to 1407):
+            Dimension of the MoE representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer decoder.
+        num_nextn_predict_layers (`int`, *optional*, defaults to 1):
+            Number of nextn predict layers in the DeepSeekV3 Model.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        n_shared_experts (`int`, *optional*, defaults to None):
+            Number of shared experts, None means dense model.
+        n_routed_experts (`int`, *optional*, defaults to None):
+            Number of routed experts, None means dense model.
+        routed_scaling_factor (`float`, *optional*, defaults to 1.0):
+            Scaling factor or routed experts.
+        topk_method (`str`, *optional*, defaults to `gready`):
+            Topk method used in routed gate.
+        n_group (`int`, *optional*, defaults to None):
+            Number of groups for routed experts.
+        topk_group (`int`, *optional*, defaults to None):
+            Number of selected groups for each token(for each token, ensuring the selected experts is only within `topk_group` groups).
+        num_experts_per_tok (`int`, *optional*, defaults to None):
+            Number of selected experts, None means dense model.
+        moe_layer_freq (`int`, *optional*, defaults to 1):
+            The frequency of the MoE layer: one expert layer for every `moe_layer_freq - 1` dense layers.
+        first_k_dense_replace (`int`, *optional*, defaults to 0):
+            Number of dense layers in shallow layers(embed->dense->dense->...->dense->moe->moe...->lm_head).
+                                                            \--k dense layers--/
+        norm_topk_prob (`bool`, *optional*, defaults to False):
+            Whether to normalize the weights of the routed experts.
+        scoring_func (`str`, *optional*, defaults to 'softmax'):
+            Method of computing expert weights.
+        aux_loss_alpha (`float`, *optional*, defaults to 0.001):
+            Auxiliary loss weight coefficient.
+        seq_aux = (`bool`, *optional*, defaults to True):
+            Whether to compute the auxiliary loss for each individual sample.
+        num_key_value_heads (`int`, *optional*):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to
+            `num_attention_heads`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 2048):
+            The maximum sequence length that this model might ever be used with.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-06):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        pad_token_id (`int`, *optional*):
+            Padding token id.
+        bos_token_id (`int`, *optional*, defaults to 1):
+            Beginning of stream token id.
+        eos_token_id (`int`, *optional*, defaults to 2):
+            End of stream token id.
+        pretraining_tp (`int`, *optional*, defaults to 1):
+            Experimental feature. Tensor parallelism rank used during pretraining. Please refer to [this
+            document](https://huggingface.co/docs/transformers/parallelism) to understand more about it. This value is
+            necessary to ensure exact reproducibility of the pretraining results. Please refer to [this
+            issue](https://github.com/pytorch/pytorch/issues/76232).
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        rope_scaling (`Dict`, *optional*):
+            Dictionary containing the scaling configuration for the RoPE embeddings. Currently supports two scaling
+            strategies: linear and dynamic. Their scaling factor must be a float greater than 1. The expected format is
+            `{"type": strategy name, "factor": scaling factor}`. When using this flag, don't update
+            `max_position_embeddings` to the expected new maximum.
+        attention_bias (`bool`, defaults to `False`, *optional*, defaults to `False`):
+            Whether to use a bias in the query, key, value and output projection layers during self-attention.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+
+    ```python
+    >>> from transformers import DeepseekV3Model, DeepseekV3Config
+
+    >>> # Initializing a Deepseek-V3 style configuration
+    >>> configuration = DeepseekV3Config()
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "deepseek_v3"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=129280,
+        hidden_size=7168,
+        intermediate_size=18432,
+        moe_intermediate_size = 2048,
+        num_hidden_layers=61,
+        num_nextn_predict_layers=1,
+        num_attention_heads=128,
+        num_key_value_heads=128,
+        n_shared_experts = 1,
+        n_routed_experts = 256,
+        ep_size = 1,
+        routed_scaling_factor = 2.5,
+        kv_lora_rank = 512,
+        q_lora_rank = 1536,
+        qk_rope_head_dim = 64,
+        v_head_dim = 128,
+        qk_nope_head_dim = 128,
+        topk_method = 'noaux_tc',
+        n_group = 8,
+        topk_group = 4,
+        num_experts_per_tok = 8,
+        moe_layer_freq = 1,
+        first_k_dense_replace = 3,
+        norm_topk_prob = True,
+        scoring_func = 'sigmoid',
+        aux_loss_alpha = 0.001,
+        seq_aux = True,
+        hidden_act="silu",
+        max_position_embeddings=4096,
+        initializer_range=0.02,
+        rms_norm_eps=1e-6,
+        use_cache=True,
+        pad_token_id=None,
+        bos_token_id=0,
+        eos_token_id=1,
+        pretraining_tp=1,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        rope_scaling=None,
+        attention_bias=False,
+        attention_dropout=0.0,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.moe_intermediate_size = moe_intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_nextn_predict_layers = num_nextn_predict_layers
+        self.num_attention_heads = num_attention_heads
+        self.n_shared_experts = n_shared_experts
+        self.n_routed_experts = n_routed_experts
+        self.ep_size = ep_size
+        self.routed_scaling_factor = routed_scaling_factor
+        self.kv_lora_rank = kv_lora_rank
+        self.q_lora_rank = q_lora_rank
+        self.qk_rope_head_dim = qk_rope_head_dim
+        self.v_head_dim = v_head_dim
+        self.qk_nope_head_dim = qk_nope_head_dim
+        self.topk_method = topk_method
+        self.n_group = n_group
+        self.topk_group = topk_group
+        self.num_experts_per_tok = num_experts_per_tok
+        self.moe_layer_freq = moe_layer_freq
+        self.first_k_dense_replace = first_k_dense_replace
+        self.norm_topk_prob = norm_topk_prob
+        self.scoring_func = scoring_func
+        self.aux_loss_alpha = aux_loss_alpha
+        self.seq_aux = seq_aux
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.pretraining_tp = pretraining_tp
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self.attention_bias = attention_bias
+        self.attention_dropout = attention_dropout
+
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )

generation_config.json

@@ -0,0 +1,4 @@
+{
+  "max_length": 131072,
+  "eos_token_id": 163586
+}

k3_headstack_recipe.json

@@ -0,0 +1,33 @@
+{
+  "base_model": "/home/ggb/mergehell/outputs/k3_fp8_experts",
+  "common_dir": "/home/ggb/mergehell/outputs/k3_headstack_rowwise_common",
+  "common_file": "model-00102-common-non-o.safetensors",
+  "config_updates": {
+    "n_shared_experts": 2,
+    "num_attention_heads": 128,
+    "num_key_value_heads": 128
+  },
+  "created_utc": "2026-06-13T20:17:22Z",
+  "expert_shards": {
+    "count": 102,
+    "mode": "relative_symlink",
+    "source": "/home/ggb/mergehell/outputs/k3_fp8_experts"
+  },
+  "logical_total_size": 3513903759664,
+  "logical_total_size_gib": 3272.577896401286,
+  "o_proj_file": "model-00103-o-proj-scale050.safetensors",
+  "o_proj_scale": 0.5,
+  "o_proj_total_size": 24192745472,
+  "o_proj_total_size_gib": 22.53125,
+  "output_dir": "/home/ggb/mergehell/outputs/k3_headstack_rowwise_o_050",
+  "rowwise_a_proj_keys": [
+    "model.layers.N.self_attn.q_a_proj.weight",
+    "model.layers.N.self_attn.kv_a_proj_with_mqa.weight"
+  ],
+  "rowwise_a_proj_merge": true,
+  "shared_output_scale": 0.5,
+  "slerp_dot_threshold": 0.9995,
+  "slerp_eps": 1e-08,
+  "slerp_t": 0.5,
+  "tensor_count": 471562
+}

k3_headstack_verify.json

@@ -0,0 +1,41 @@
+{
+  "apparent_entry_bytes": 24244741638,
+  "expert_symlink_count": 102,
+  "files_in_index": 104,
+  "shape_checks": {
+    "model.layers.1.mlp.shared_experts.down_proj.weight": [
+      7168,
+      4096
+    ],
+    "model.layers.1.mlp.shared_experts.gate_proj.weight": [
+      4096,
+      7168
+    ],
+    "model.layers.1.mlp.shared_experts.up_proj.weight": [
+      4096,
+      7168
+    ],
+    "model.layers.1.self_attn.kv_a_proj_with_mqa.weight": [
+      576,
+      7168
+    ],
+    "model.layers.1.self_attn.kv_b_proj.weight": [
+      32768,
+      512
+    ],
+    "model.layers.1.self_attn.o_proj.weight": [
+      7168,
+      16384
+    ],
+    "model.layers.1.self_attn.q_a_proj.weight": [
+      1536,
+      7168
+    ],
+    "model.layers.1.self_attn.q_b_proj.weight": [
+      24576,
+      1536
+    ]
+  },
+  "tensor_count": 471562,
+  "total_symlink_count": 103
+}

model-00003-layer004-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab6cb887de18bc3f7a10552d1a68d3a2ba89db21a9fe50fabf058045952b6711
+size 33831718248

model-00005-layer006-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c1cb5ca25879f8e38df48c52d385de205f3ed1e86b76d46dbe34e143c7bc7d6
+size 33831718248

model-00007-layer008-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1599376ce4da71da9b42e9dac700c1c114fcab6e4871b560c2384e8d95017416
+size 33831718248

model-00008-layer009-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:968c7e554be42645c814e659d6e0b2d0c78236ed836317fc1b4fbeae4bf5dce6
+size 33831718248

model-00010-layer011-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cea2ab3a68cc124c12fa6de8ba516b86f57cb96098ae825883fa8498c24cf3af
+size 33831722856

model-00011-layer012-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0411880abbb4b7868a1205f42f51031d1b62a719d7322749912e8682d0a7822b
+size 33831722856

model-00012-layer013-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcca0cfd29ff21dd02c1fb8c7d6cce630750b760df7f019e621476fec159403f
+size 33831722856

model-00014-layer015-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b27fff951a1ebc08b5685f08ea97648fcde89e5d5184013311920cd63f60eb27
+size 33831722856

model-00016-layer017-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed7904cac9949215ad836d31bd5bc3f52d4fc1acd9964be6202f4ed27b1b1dfd
+size 33831722856

model-00017-layer018-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f028f09c7f599624d1ce7a88aa646beda66e8cf665439738862cfeea85ffc42
+size 33831722856

model-00018-layer019-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5af5206a475f21693f2a314fedb805f4e779b32cdc62c5747bae728b8299aa8
+size 33831722856

model-00020-layer021-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aed8c32a2c9e30cf1b945530d371b15d7e1ef6039c10961653fa3af9289f0fde
+size 33831722856

model-00024-layer025-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:643432f754d37488eadc2cdc646bdc80c41143406e68676f16a1eca5654c0e2a
+size 33831722856

model-00029-layer030-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9c25fd7677d5b8606f0e88d87cbab82b3e89143e29e0656c4164597807eb859
+size 33831722856

model-00030-layer031-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7efda5a37fa3baf35879c9b3660b7651d8010be3ab989b883f242f0e053b769d
+size 33831722856

model-00031-layer032-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f960e7c19e1a6843c43ae19e92d5de627a495bb432959e810519528395c8e3ba
+size 33831722856

model-00033-layer034-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd6f9dc29a0a618b24c9cb66d857cc559f84df374f5ec90021fac7c28cc099f9
+size 33831722856

model-00034-layer035-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a5e5bf55f3d2623d0f6080f25bf87902741514e75464605df9a452920d1ab30
+size 33831722856

model-00038-layer039-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41f30c39d802182107b504d3a53c9110f8a7e6d2ba4e90618d8605c81166ec5e
+size 33831722856

model-00039-layer040-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb5fc29ea1f60e74bb418bb71b137853e72b3221658881c925bef7914b729ef1
+size 33831722856

model-00042-layer043-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d2956c625801fb552381009db2b996ea57da5630bb7c1dee67562e34f8f9910
+size 33831722856

model-00044-layer045-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7083acb097db0c79fe0eb7618eac389fc57902ae8352ab71379b050417cd8c2e
+size 33831722856

model-00047-layer048-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:165edec1f5efcf3d2e58fc864a6cdfee66236f324f9ac94c53be13e0f828f7af
+size 33831722856

model-00052-layer053-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f6ed83ffda28e608edd0cfb91f595b10236a8972736023c0ccfec8efdfa7ace
+size 33831722856

model-00053-layer054-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7882919d8abde1282e5cb2db7940157c538600306b08be25894e1ac27adc09b8
+size 33831722856

model-00055-layer056-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8649684b3f8472aa13c1f5105e979475a5f73eeb424f49f401a80f3b301f2152
+size 33831722856

model-00059-layer060-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dacbdbac98d354269d94973944e4fc2cfcb8229327d146cb767d7f75da475bf9
+size 33831722856

model-00061-layer062-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f886477a0fc03624cad9958b635dd1dbd9582b3acdfc364909a4baa6e87c24d
+size 33831722856

model-00064-layer065-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7df77c77f008c3397e2a98cbcfe764201834a5fbac4d99ba50acb5963aa7714a
+size 33831722856

model-00066-layer067-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44f463f101b6562d9ef0966b7a155020fdaea490d2f06b3723b0e63e8dea269c
+size 33831722856

model-00067-layer068-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a286c13e986e4263d66d265b47b2408b6d1811cbf14c83e313caf79a35e545dd
+size 33831722856

model-00073-layer074-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e6f1694c6eed798bf4de6bdeba73bf548cd70ceff235a26e028024bb6f76ee9
+size 33831722856

model-00080-layer081-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aec66655195d3d48b31d05f5db762781205101f381dff1886ad434a0a4451312
+size 33831722856

model-00086-layer087-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d76ed31e2e0a3d7c69beecfa584b4ce05e776740676ac43d163e42507dd75e8
+size 33831722856

model-00090-layer091-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f734838a9d8dcda7b3b78db14f054849942832927c4e025ec984e1f130c1c8b
+size 33831722856

model-00094-layer095-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c66ee0305d2705f2043f8a3323ced88979252329ac32318250e0f6020a7a9021
+size 33831722856

model-00096-layer097-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e809dedce5e03b47fb8b97781ac0285072b92af2fab3a6618f6f43f55115389b
+size 33831722856

model-00098-layer099-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4dbcd120d38087c33c8429aecd25af423541b3a479bbbce8c99c0486067e9a4
+size 33831722856

model-00099-layer100-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8578be7bfbc9cf1daaf2ef0b4edecaec44421b5b1114a3783fbc81e13c7dcc3
+size 33831727464

model-00100-layer101-experts.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:406b350372893fcd81f6e81f7d9d2761d876c2402e0242849ab71df0d1f85088
+size 33831727464

model.safetensors.index.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4d5eea83cd1e5218418ecc849be0c3a1865b2e963ee5bcbe6f03a7ed730f347
+size 49086552

tiktoken.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6c497a7469b33ced9c38afb1ad6e47f03f5e5dc05f15930799210ec050c5103
+size 2795286

tokenization_kimi.py

@@ -0,0 +1,323 @@
+import os
+import tiktoken
+
+from logging import getLogger
+from pathlib import Path
+from typing import (
+    cast,
+    Tuple,
+    Dict,
+    Iterator,
+    List,
+    Union,
+    Optional,
+)
+from shutil import copyfile
+from tiktoken.load import load_tiktoken_bpe
+from tokenizers import AddedToken, pre_tokenizers, Regex
+from transformers.tokenization_utils import PreTrainedTokenizer
+from transformers.models.gpt2.tokenization_gpt2 import bytes_to_unicode
+
+
+
+logger = getLogger(__name__)
+VOCAB_FILES_NAMES = {"vocab_file": "tiktoken.model"}
+
+class TikTokenTokenizer(PreTrainedTokenizer):
+    """
+    Tokenizing and encoding/decoding text using the Tiktoken tokenizer. See megatron/tokenizer/tiktoken_tokenizer.py.
+
+    This tokenizer inherits from [`PreTrainedTokenizer`] which contains most of the main methods. Users should refer to
+    this superclass for more information regarding those methods.
+
+    Args:
+        vocab_file (`str`):
+            The path to the Tiktoken model file.
+        bos_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"<|begin_of_text|>",`):
+            The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token.
+        eos_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"<|end_of_text|>"`):
+            The end of sequence token.
+        unk_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"<|reserved_special_token_249|>"`):
+            The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
+            token instead. The second to last item in special_tokens.
+        pad_token (`str` or `tokenizers.AddedToken`, *optional*, defaults to `"<|reserved_special_token_250|>"`):
+            The token used for padding, for example when batching sequences of different lengths.
+        additional_special_tokens (list of `str`, *optional*):
+            A tuple or a list of additional tokens, which will be marked as `special`, meaning that they will be
+            skipped when decoding if `skip_special_tokens` is set to `True`.
+    """
+
+    vocab_files_names = VOCAB_FILES_NAMES
+
+    model_input_names = ["input_ids", "attention_mask"]
+
+    special_tokens: Dict[str, int]
+
+    num_reserved_special_tokens = 256
+
+    pat_str = "|".join(
+        [
+            r"""[\p{Han}]+""",
+            r"""[^\r\n\p{L}\p{N}]?[\p{Lu}\p{Lt}\p{Lm}\p{Lo}\p{M}&&[^\p{Han}]]*[\p{Ll}\p{Lm}\p{Lo}\p{M}&&[^\p{Han}]]+(?i:'s|'t|'re|'ve|'m|'ll|'d)?""",
+            r"""[^\r\n\p{L}\p{N}]?[\p{Lu}\p{Lt}\p{Lm}\p{Lo}\p{M}&&[^\p{Han}]]+[\p{Ll}\p{Lm}\p{Lo}\p{M}&&[^\p{Han}]]*(?i:'s|'t|'re|'ve|'m|'ll|'d)?""",
+            r"""\p{N}{1,3}""",
+            r""" ?[^\s\p{L}\p{N}]+[\r\n]*""",
+            r"""\s*[\r\n]+""",
+            r"""\s+(?!\S)""",
+            r"""\s+""",
+        ]
+    )
+
+    def __init__(
+        self,
+        vocab_file,
+        bos_token: Union[str, AddedToken]="[BOS]",
+        eos_token: Union[str, AddedToken]="[EOS]",
+        unk_token: Union[str, AddedToken, None]=None,
+        pad_token: Union[str, AddedToken, None]=None,
+        additional_special_tokens: List[str]=None,
+        added_tokens_decoder: Optional[dict] = None,
+        **kwargs,
+    ):
+        assert os.path.isfile(vocab_file), vocab_file
+
+        if additional_special_tokens is None:
+            additional_special_tokens  = [
+                "<|im_end|>",
+                "<|im_user|>", 
+                "<|im_assistant|>", 
+                "<|start_header_id|>", 
+                "<|end_header_id|>", 
+                "[EOT]", 
+                "<|im_system|>", 
+                "<|im_middle|>",
+            ] 
+        
+        special_tokens_mapping = {
+        i: added_tokens_decoder[i].content for i in added_tokens_decoder
+    }
+
+        self.vocab_file = vocab_file
+        mergeable_ranks = load_tiktoken_bpe(vocab_file)
+        num_base_tokens = len(mergeable_ranks)
+        self.special_tokens = {
+            special_tokens_mapping.get(i, f"<|reserved_token_{i}|>"): i
+            for i in range(
+                num_base_tokens, num_base_tokens + self.num_reserved_special_tokens + 2
+            )
+        }
+
+
+
+        self.model = tiktoken.Encoding(
+            name=Path(vocab_file).name,
+            pat_str=self.pat_str,
+            mergeable_ranks=mergeable_ranks,
+            special_tokens=self.special_tokens,
+        )
+        logger.info(f"Reloaded tiktoken model from {vocab_file}")
+
+        self.n_words: int = self.model.n_vocab
+        # BOS / EOS token IDs
+        self.bos_id: int = self.special_tokens[str(bos_token)]
+        self.eos_id: int = self.special_tokens[str(eos_token)]
+        logger.info(
+            f"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}"
+        )
+
+        self.pad_id: int = self.special_tokens[str(pad_token)]
+        self.unk_id: int = self.special_tokens[str(unk_token)]
+
+        self.byte_encoder = bytes_to_unicode()
+        self.byte_decoder = {v: k for k, v in self.byte_encoder.items()}
+
+        self.decoder = {}
+        for i in range(self.n_words):
+            # Taken from https://gist.github.com/xenova/a452a6474428de0182b17605a98631ee
+            decoding = ''.join([
+                self.byte_encoder[ord(char)] for char in
+                self.model.decode_single_token_bytes(i).decode('latin-1')
+            ])
+            self.decoder[i] = decoding
+
+        self.encoder = {}
+        for i in range(self.n_words):
+            if i in self.decoder:
+                self.encoder[self.decoder[i]] = i
+
+        super().__init__(
+            bos_token=bos_token,
+            eos_token=eos_token,
+            unk_token=unk_token,
+            pad_token=pad_token,
+            additional_special_tokens=additional_special_tokens,
+            **kwargs,
+        )
+        self.all_special_ids_set = set(self.all_special_ids)
+
+    def encode(
+        self,
+        text: str,
+        allow_special_tokens: bool = True,
+        **kwargs
+    ) -> List[int]:
+        """
+        Encodes a string into a list of token IDs.
+
+        Args:
+            text (str): The input string to be encoded.
+
+        Returns:
+            list[int]: A list of token IDs.
+        """
+        # If there are other args, we should call super().encode because there are a lot of code
+        # to handle those args. supper().encode finally will call _tokenize and _convert_token_to_id.
+        # NOTE: our encode method is not compatible with the super().encode method,
+        #   e.g. split_special_tokens' default is True in our encode method.
+        if len(kwargs) > 0:
+            logger.warning( f"Calling super().encode with {kwargs}" )
+            return super().encode(text, **kwargs)
+
+        assert type(text) is str
+
+        # The tiktoken tokenizer can handle <=400k chars without
+        # pyo3_runtime.PanicException.
+        TIKTOKEN_MAX_ENCODE_CHARS = 400_000
+
+        # https://github.com/openai/tiktoken/issues/195
+        # Here we iterate over subsequences and split if we exceed the limit
+        # of max consecutive non-whitespace or whitespace characters.
+        MAX_NO_WHITESPACES_CHARS = 25_000
+
+        texts = self.pre_tokenizer_process(text)
+
+        all_substrs = []
+        for text in texts:
+            substrs = (
+                substr
+                for i in range(0, len(text), TIKTOKEN_MAX_ENCODE_CHARS)
+                for substr in self._split_whitespaces_or_nonwhitespaces(
+                    text[i: i + TIKTOKEN_MAX_ENCODE_CHARS], MAX_NO_WHITESPACES_CHARS
+                )
+            )
+            all_substrs.extend(substrs)
+
+        t: List[int] = []
+        for substr in all_substrs:
+            if allow_special_tokens:
+                t.extend(
+                    # we should consider special token as a common token
+                    self.model.encode(
+                        substr,
+                        allowed_special="all",
+                    )
+                )
+            else:
+                t.extend(
+                    # we should consider special token as a common token
+                    self.model.encode(
+                        substr,
+                        disallowed_special=(),
+                    )
+                )
+
+        return t
+
+    def decode(
+        self,
+        token_ids: Union[int, List[int]],
+        **kwargs
+    ) -> str:
+        """
+        Decodes a list of token IDs into a string.
+
+        Args:
+            token_ids (List[int]): The list of token IDs to be decoded.
+
+        Returns:
+            str: The decoded string.
+        """
+        # If there are other args, we should call super().decode because there are a lot of code
+        # to handle those args. supper().encode finally will call convert_tokens_to_string and _convert_id_to_token.
+        if len(kwargs) > 0:
+            return super().decode(token_ids, **kwargs)
+
+        if type(token_ids) is int:
+            token_ids = [token_ids]
+
+        return self.model.decode(cast(List[int], token_ids))
+
+    @staticmethod
+    def _split_whitespaces_or_nonwhitespaces(
+        s: str, max_consecutive_slice_len: int
+    ) -> Iterator[str]:
+        """
+        Splits the string `s` so that each substring contains no more than `max_consecutive_slice_len`
+        consecutive whitespaces or consecutive non-whitespaces.
+        """
+        current_slice_len = 0
+        current_slice_is_space = s[0].isspace() if len(s) > 0 else False
+        slice_start = 0
+
+        for i in range(len(s)):
+            is_now_space = s[i].isspace()
+
+            if current_slice_is_space ^ is_now_space:
+                current_slice_len = 1
+                current_slice_is_space = is_now_space
+            else:
+                current_slice_len += 1
+                if current_slice_len > max_consecutive_slice_len:
+                    yield s[slice_start:i]
+                    slice_start = i
+                    current_slice_len = 1
+        yield s[slice_start:]
+
+    def pre_tokenizer_process(self, text: str) -> List[str]:
+        """
+        pre-tokenizes the input text into a list of tokens.
+        This method is used to split the input text into smaller chunks for internal processing.
+        """
+        return [text]
+
+
+    """ ----- Below are the abstract methods required by PreTrainedTokenizer ----- """
+    @property
+    def vocab_size(self) -> int:
+        return self.n_words
+
+    def get_vocab(self) -> Dict[str, int]:
+        return self.encoder
+
+    def _tokenize(self, text: str, **kwargs) -> List[str]:
+        return [
+            self.decoder[t]
+            for t in self.encode(text)
+        ]
+
+    def _convert_token_to_id(self, token: str) -> int:
+        return self.encoder.get(token, self.unk_id)
+
+    def _convert_id_to_token(self, index: int) -> str:
+        return self.decoder.get(index)
+
+    @staticmethod
+    def clean_up_tokenization(out_string: str) -> str:
+        return out_string
+
+    def convert_tokens_to_string(self, tokens: List[str]) -> str:
+        text = ''.join(tokens)
+        text = bytearray([self.byte_decoder[c] for c in text]).decode('utf-8', 'replace')
+        return text
+
+    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
+        if not os.path.isdir(save_directory):
+            raise ValueError(f"vocabulary path ({save_directory}) should be a directory")
+        out_vocab_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
+        )
+
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
+            copyfile(self.vocab_file, out_vocab_file)
+
+        return (out_vocab_file,)