Version1 (#1)

- upload mxfp4 deepseek-v3 (0c7c716272255bca2a02e7952d747819ad980317)

chat_template.jinja

@@ -0,0 +1,14 @@
+{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='', is_first_sp=true, is_last_user=false) %}{%- for message in messages %}{%- if message['role'] == 'system' %}{%- if ns.is_first_sp %}{% set ns.system_prompt = ns.system_prompt + message['content'] %}{% set ns.is_first_sp = false %}{%- else %}{% set ns.system_prompt = ns.system_prompt + '
+
+' + message['content'] %}{%- endif %}{%- endif %}{%- endfor %}{{ bos_token }}{{ ns.system_prompt }}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{%- set ns.is_first = false -%}{%- set ns.is_last_user = true -%}{{'<｜User｜>' + message['content'] + '<｜Assistant｜>'}}{%- endif %}{%- if message['role'] == 'assistant' and message['tool_calls'] is defined and message['tool_calls'] is not none %}{%- set ns.is_last_user = false -%}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{%- endif %}{%- set ns.is_first = false %}{%- set ns.is_tool = false -%}{%- set ns.is_output_first = true %}{%- for tool in message['tool_calls'] %}{%- if not ns.is_first %}{%- if message['content'] is none %}{{'<｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '
+' + '```json' + '
+' + tool['function']['arguments'] + '
+' + '```' + '<｜tool▁call▁end｜>'}}{%- else %}{{message['content'] + '<｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '
+' + '```json' + '
+' + tool['function']['arguments'] + '
+' + '```' + '<｜tool▁call▁end｜>'}}{%- endif %}{%- set ns.is_first = true -%}{%- else %}{{'
+' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '
+' + '```json' + '
+' + tool['function']['arguments'] + '
+' + '```' + '<｜tool▁call▁end｜>'}}{%- endif %}{%- endfor %}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- if message['role'] == 'assistant' and (message['tool_calls'] is not defined or message['tool_calls'] is none)%}{%- set ns.is_last_user = false -%}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{% set content = message['content'] %}{{content + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_last_user = false -%}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'
+<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_last_user and not ns.is_tool %}{{'<｜Assistant｜>'}}{% endif %}

config.json

@@ -0,0 +1,452 @@
+{
+  "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,
+  "bos_token_id": 0,
+  "eos_token_id": 1,
+  "ep_size": 1,
+  "first_k_dense_replace": 3,
+  "hidden_act": "silu",
+  "hidden_size": 7168,
+  "initializer_range": 0.02,
+  "intermediate_size": 18432,
+  "kv_lora_rank": 512,
+  "max_position_embeddings": 163840,
+  "model_type": "deepseek_v3",
+  "moe_intermediate_size": 2048,
+  "moe_layer_freq": 1,
+  "n_group": 8,
+  "n_routed_experts": 256,
+  "n_shared_experts": 1,
+  "norm_topk_prob": true,
+  "num_attention_heads": 128,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 61,
+  "num_key_value_heads": 128,
+  "num_nextn_predict_layers": 1,
+  "pretraining_tp": 1,
+  "q_lora_rank": 1536,
+  "qk_nope_head_dim": 128,
+  "qk_rope_head_dim": 64,
+  "quantization_config": {
+    "algo_config": [
+      {
+        "compute_scale_loss": "MAE",
+        "model_decoder_layers": "model.layers",
+        "name": "autosmoothquant",
+        "scaling_layers": [
+          {
+            "inp": "mlp.gate_proj",
+            "layers": [
+              "mlp.gate_proj",
+              "mlp.up_proj"
+            ],
+            "module2inspect": "mlp",
+            "prev_op": "post_attention_layernorm"
+          },
+          {
+            "inp": "mlp.down_proj",
+            "layers": [
+              "mlp.down_proj"
+            ],
+            "prev_op": "mlp.up_proj"
+          },
+          {
+            "inp": "down_proj",
+            "layers": [
+              "down_proj"
+            ],
+            "prev_op": "up_proj"
+          }
+        ]
+      }
+    ],
+    "exclude": [
+      "model.layers.0.self_attn.q_a_proj",
+      "model.layers.0.self_attn.q_b_proj",
+      "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.1.self_attn.q_a_proj",
+      "model.layers.1.self_attn.q_b_proj",
+      "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.2.self_attn.q_a_proj",
+      "model.layers.2.self_attn.q_b_proj",
+      "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.3.self_attn.q_a_proj",
+      "model.layers.3.self_attn.q_b_proj",
+      "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.4.self_attn.q_a_proj",
+      "model.layers.4.self_attn.q_b_proj",
+      "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.5.self_attn.q_a_proj",
+      "model.layers.5.self_attn.q_b_proj",
+      "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.6.self_attn.q_a_proj",
+      "model.layers.6.self_attn.q_b_proj",
+      "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.7.self_attn.q_a_proj",
+      "model.layers.7.self_attn.q_b_proj",
+      "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.8.self_attn.q_a_proj",
+      "model.layers.8.self_attn.q_b_proj",
+      "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.9.self_attn.q_a_proj",
+      "model.layers.9.self_attn.q_b_proj",
+      "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.10.self_attn.q_a_proj",
+      "model.layers.10.self_attn.q_b_proj",
+      "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.11.self_attn.q_a_proj",
+      "model.layers.11.self_attn.q_b_proj",
+      "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.12.self_attn.q_a_proj",
+      "model.layers.12.self_attn.q_b_proj",
+      "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.13.self_attn.q_a_proj",
+      "model.layers.13.self_attn.q_b_proj",
+      "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.14.self_attn.q_a_proj",
+      "model.layers.14.self_attn.q_b_proj",
+      "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.15.self_attn.q_a_proj",
+      "model.layers.15.self_attn.q_b_proj",
+      "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.16.self_attn.q_a_proj",
+      "model.layers.16.self_attn.q_b_proj",
+      "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.17.self_attn.q_a_proj",
+      "model.layers.17.self_attn.q_b_proj",
+      "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.18.self_attn.q_a_proj",
+      "model.layers.18.self_attn.q_b_proj",
+      "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.19.self_attn.q_a_proj",
+      "model.layers.19.self_attn.q_b_proj",
+      "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.20.self_attn.q_a_proj",
+      "model.layers.20.self_attn.q_b_proj",
+      "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.21.self_attn.q_a_proj",
+      "model.layers.21.self_attn.q_b_proj",
+      "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.22.self_attn.q_a_proj",
+      "model.layers.22.self_attn.q_b_proj",
+      "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.23.self_attn.q_a_proj",
+      "model.layers.23.self_attn.q_b_proj",
+      "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.24.self_attn.q_a_proj",
+      "model.layers.24.self_attn.q_b_proj",
+      "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.25.self_attn.q_a_proj",
+      "model.layers.25.self_attn.q_b_proj",
+      "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.26.self_attn.q_a_proj",
+      "model.layers.26.self_attn.q_b_proj",
+      "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.27.self_attn.q_a_proj",
+      "model.layers.27.self_attn.q_b_proj",
+      "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.28.self_attn.q_a_proj",
+      "model.layers.28.self_attn.q_b_proj",
+      "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.29.self_attn.q_a_proj",
+      "model.layers.29.self_attn.q_b_proj",
+      "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.30.self_attn.q_a_proj",
+      "model.layers.30.self_attn.q_b_proj",
+      "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.31.self_attn.q_a_proj",
+      "model.layers.31.self_attn.q_b_proj",
+      "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.32.self_attn.q_a_proj",
+      "model.layers.32.self_attn.q_b_proj",
+      "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.33.self_attn.q_a_proj",
+      "model.layers.33.self_attn.q_b_proj",
+      "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.34.self_attn.q_a_proj",
+      "model.layers.34.self_attn.q_b_proj",
+      "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.35.self_attn.q_a_proj",
+      "model.layers.35.self_attn.q_b_proj",
+      "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.36.self_attn.q_a_proj",
+      "model.layers.36.self_attn.q_b_proj",
+      "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.37.self_attn.q_a_proj",
+      "model.layers.37.self_attn.q_b_proj",
+      "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.38.self_attn.q_a_proj",
+      "model.layers.38.self_attn.q_b_proj",
+      "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.39.self_attn.q_a_proj",
+      "model.layers.39.self_attn.q_b_proj",
+      "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.40.self_attn.q_a_proj",
+      "model.layers.40.self_attn.q_b_proj",
+      "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.41.self_attn.q_a_proj",
+      "model.layers.41.self_attn.q_b_proj",
+      "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.42.self_attn.q_a_proj",
+      "model.layers.42.self_attn.q_b_proj",
+      "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.43.self_attn.q_a_proj",
+      "model.layers.43.self_attn.q_b_proj",
+      "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.44.self_attn.q_a_proj",
+      "model.layers.44.self_attn.q_b_proj",
+      "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.45.self_attn.q_a_proj",
+      "model.layers.45.self_attn.q_b_proj",
+      "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.46.self_attn.q_a_proj",
+      "model.layers.46.self_attn.q_b_proj",
+      "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.47.self_attn.q_a_proj",
+      "model.layers.47.self_attn.q_b_proj",
+      "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.48.self_attn.q_a_proj",
+      "model.layers.48.self_attn.q_b_proj",
+      "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.49.self_attn.q_a_proj",
+      "model.layers.49.self_attn.q_b_proj",
+      "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.50.self_attn.q_a_proj",
+      "model.layers.50.self_attn.q_b_proj",
+      "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.51.self_attn.q_a_proj",
+      "model.layers.51.self_attn.q_b_proj",
+      "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.52.self_attn.q_a_proj",
+      "model.layers.52.self_attn.q_b_proj",
+      "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.53.self_attn.q_a_proj",
+      "model.layers.53.self_attn.q_b_proj",
+      "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.54.self_attn.q_a_proj",
+      "model.layers.54.self_attn.q_b_proj",
+      "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.55.self_attn.q_a_proj",
+      "model.layers.55.self_attn.q_b_proj",
+      "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.56.self_attn.q_a_proj",
+      "model.layers.56.self_attn.q_b_proj",
+      "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.57.self_attn.q_a_proj",
+      "model.layers.57.self_attn.q_b_proj",
+      "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.58.self_attn.q_a_proj",
+      "model.layers.58.self_attn.q_b_proj",
+      "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.59.self_attn.q_a_proj",
+      "model.layers.59.self_attn.q_b_proj",
+      "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.60.self_attn.q_a_proj",
+      "model.layers.60.self_attn.q_b_proj",
+      "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",
+      "lm_head"
+    ],
+    "export": {
+      "kv_cache_group": [],
+      "min_kv_scale": 0.0,
+      "pack_method": "reorder",
+      "weight_format": "real_quantized",
+      "weight_merge_groups": null
+    },
+    "global_quant_config": {
+      "bias": null,
+      "input_tensors": {
+        "ch_axis": -1,
+        "dtype": "fp4",
+        "group_size": 32,
+        "is_dynamic": true,
+        "is_scale_quant": false,
+        "mx_element_dtype": null,
+        "observer_cls": "PerBlockMXObserver",
+        "qscheme": "per_group",
+        "round_method": "half_even",
+        "scale_calculation_mode": "even",
+        "scale_format": "e8m0",
+        "scale_type": "float",
+        "symmetric": null
+      },
+      "output_tensors": null,
+      "target_device": null,
+      "weight": {
+        "ch_axis": -1,
+        "dtype": "fp4",
+        "group_size": 32,
+        "is_dynamic": false,
+        "is_scale_quant": false,
+        "mx_element_dtype": null,
+        "observer_cls": "PerBlockMXObserver",
+        "qscheme": "per_group",
+        "round_method": "half_even",
+        "scale_calculation_mode": "even",
+        "scale_format": "e8m0",
+        "scale_type": "float",
+        "symmetric": null
+      }
+    },
+    "kv_cache_quant_config": {},
+    "layer_quant_config": {},
+    "layer_type_quant_config": {},
+    "quant_method": "quark",
+    "quant_mode": "eager_mode",
+    "softmax_quant_spec": null,
+    "version": "0.10+d757f3f3e0"
+  },
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": {
+    "beta_fast": 32,
+    "beta_slow": 1,
+    "factor": 40,
+    "mscale": 1.0,
+    "mscale_all_dim": 1.0,
+    "original_max_position_embeddings": 4096,
+    "type": "yarn"
+  },
+  "rope_theta": 10000,
+  "routed_scaling_factor": 2.5,
+  "scoring_func": "sigmoid",
+  "seq_aux": true,
+  "tie_word_embeddings": false,
+  "topk_group": 4,
+  "topk_method": "noaux_tc",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.53.2",
+  "use_cache": true,
+  "v_head_dim": 128,
+  "vocab_size": 129280
+}

configuration_deepseek.py

@@ -0,0 +1,210 @@
+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 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 0,
+  "eos_token_id": 1,
+  "transformers_version": "4.53.2"
+}

model-00001-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f778638be1f70d723f39f8d06416fc0277a556def7f4290c31a579bbd215927
+size 4995884768

model-00002-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf90557853cc8f54889dc60cad1c2b8bda4c059481a299825dc4fbf4e3c7fd77
+size 4995052496

model-00003-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2dbfb5bd3a393b94cc02e61147d15c9a6ec296592a1240ecd2eb54704149a7e
+size 4999071368

model-00004-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d5e786f3f08201620f0ad6c447e98719965fbec4991c3fe873ca9f7117c94fc
+size 4994943096

model-00005-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:899d0dcd20f37432d03f2307e38a9d2c4b7021ed59cb34ad72c1c4a442746ccf
+size 4994943232

model-00006-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa9ee4ab184279c388525635cdefdabccf486d34cfdd924f06376f648647f329
+size 4994943456

model-00007-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d9b5c77354958773fabb9ce15c86aa26dbd30350c385daa61466fbc0e20dbdd
+size 4999180624

model-00008-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a341f6029020f9d561f95350dfd37b2894f239ddf7cd2b67e14ccff839ac74ad
+size 4994943096

model-00009-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd57dce52649b06964a27e795df8fbe4b11f9cd3f26811a94a730d3f6c95a6b3
+size 4994943096

model-00010-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44ee08146393c389692df3fe717e3baf0d3fa5ca8dbde21ae184486f7a818542
+size 4994943720

model-00011-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9937f7d352e4cf22793923cc7952f049eb0c0480ba936946fac33ea7120cf9cc
+size 4916044152

model-00012-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57fef23f2ae5b6e10e6502d016ef26790472d7414ae8def07342aa98687757f9
+size 5000091936

model-00013-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d923b18f6667b62ebdb4f54afc1d8bd9bec365378617855cfb2048bfbb8243b7
+size 4994944296

model-00014-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb531c293e31f51ae520e51481ede8a04fb5a73fc433e17ef07268b83f4b0b0a
+size 4994944392

model-00015-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91ba4f56d3a71deef0b16b48794a754ea6c0398caa00603c4cbfa3b0ba7e4cff
+size 4994944648

model-00016-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1847dfb119d6ce1803d715b4eb1c95f916af1a657fb2be799a89e3ebfccb9698
+size 4999181952

model-00017-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48671bf382c2ed1d952cb15526e35c1ed04fa4554483c3db36dc44c5f8fc996b
+size 4994944288

model-00018-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:474fa2a5eb81e632a9dcd1d9e3071e85ad5fb51bb203e90c1502463301b89380
+size 4994944296

model-00019-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dde9bc93227e80d1210ce126977f86c3f1b48bc99b35a6485a4139c1eb862056
+size 4994944536

model-00020-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4249fc283d60f7c5727781aa77cce363a0f8ebacbb0b83125d366ae50c952ead
+size 4767862480

model-00021-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62c0c1a130eb729eab491e8bc4c46ca7c796e9e0fb155f7ecf5e3e084b1c8783
+size 5000091936

model-00022-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb3e028c10344770cfcb020c1b7c2d3053743ad15f9f297718fc9176a1a22f14
+size 4994944296

model-00023-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a20bd7b41f3238c46a9d9d872883c0a3180b9f79b2fd9c85e8f858300565ade
+size 4994944392

model-00024-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e2261c65a19665332be3dcb8e8f322fa62d6198e6c6a41d77fea41ea2ada3b2
+size 4994944648

model-00025-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d421f67ac55b85a7e8d02679a654d1676e1aa26a09465489642df63419537870
+size 4999181952

model-00026-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:185403d53bfc61c8806efc65d35bf7f88b9ab9eebf11fde869b2d1a10726a043
+size 4994944288

model-00027-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:375af68c1da3733821f38049cf39f602b753cb41f777aaca07309468e5caeac0
+size 4994944296

model-00028-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65210438552cb36f2c954c5812987b7545a5a9f88930d3d569bf6c9f864e5629
+size 4994944536

model-00029-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e654e9ae8c58efb0089ab1b9c49de607b24644365dd520da016f991f320a1203
+size 4767862480

model-00030-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45b8db2d004cb1806bd9a6b5d4dec4c82133b035999a01ff907c0ae4f14a4b2b
+size 5000091936

model-00031-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc348bd54630faacd6f7e4f32a36e99d66640b9d84ddd59d99ba14c0d734d9a9
+size 4994944296

model-00032-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffdc7abe493ab218d163d498aa6f2332826470cba5ced44ec7206f7f4eb1a18f
+size 4994944392

model-00033-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5072926a14a9210a3dbbff4824012ee7f72e45ea8f2029434a6558333dbeeb66
+size 4994944648

model-00034-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4aaba166a2355a804c565a77f1f0384b82e1566d05c0910a928db93f2d32adf
+size 4999181952

model-00035-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2184d9b1ef4e82ae82609a1cbd7b92c7d83c69fc42c9749e93ca51464f329b2c
+size 4994944288

model-00036-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c74b3b49cc4637777115b3d3bd9aedf91102e2413ab375a5265a5543f078ae98
+size 4994944296

model-00037-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e9992c51438f70befaf876ff7c5b29ae8ebd413758b8c9dc5ac3d8130726f01
+size 4994944536

model-00038-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7dc244be1b918a896be399e86c1738d8c62f18c9fd04010be2531643c2b1571
+size 4767862480

model-00039-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30daa8ec577fb405af3d481ff4479764f024c495a82598b6740eaebc44b6aaea
+size 5000091936

model-00040-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21da904f8f8682bdafaf3efc9197648368e02a73f77e8315c1f819ce37382016
+size 4994944296

model-00041-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6590565c74235c757ba7ac9c0f9d98eda5ee647d8968eeb096a30de47a53bf71
+size 4994944392

model-00042-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27620f4c97f92604ed10dd766afaecd4cb4d81082bb4f7bd46c5a6d903dbd650
+size 4994944648

model-00043-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ed07ef726be49e982a748a7c6d798f017f0678c2b421eda59fd0ba70e44b58f
+size 4999181952

model-00044-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de12ec03f4f1c74f1258e2a060e941e1a2b5d918aa49ed9124a7ef79f985f254
+size 4994944288

model-00045-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1bea77e71a772046628590c8599bde92e7b5b84278dc3d9604264ab8243b215
+size 4994944296

model-00046-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:992ecd8324e44ac79d35207d27b6c2324e15a9b30ccf8292af836e6d03b65e5d
+size 4994944536