upload 0528 moe quantized model

config.json

@@ -0,0 +1,419 @@
+{
+  "architectures": [
+    "DeepseekV3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_deepseek.DeepseekV3Config",
+    "AutoModel": "modeling_deepseek.DeepseekV3Model",
+    "AutoModelForCausalLM": "modeling_deepseek.DeepseekV3ForCausalLM"
+  },
+  "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,
+  "pad_token_id": 2,
+  "q_lora_rank": 1536,
+  "qk_nope_head_dim": 128,
+  "qk_rope_head_dim": 64,
+  "quantization_config": {
+    "algo_config": null,
+    "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
+      }
+    },
+    "layer_quant_config": {},
+    "layer_type_quant_config": {},
+    "quant_method": "quark",
+    "quant_mode": "eager_mode",
+    "softmax_quant_spec": null,
+    "version": "0.10+577c94ac99"
+  },
+  "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",
+  "tie_word_embeddings": false,
+  "topk_group": 4,
+  "topk_method": "noaux_tc",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "unsloth_fixed": true,
+  "use_cache": true,
+  "v_head_dim": 128,
+  "vocab_size": 129280
+}

configuration_deepseek.py

@@ -0,0 +1,199 @@
+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.
+        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',
+        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,
+        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
+        # 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.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 +1,9 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 0,
+  "do_sample": true,
+  "eos_token_id": 1,
+  "temperature": 0.6,
+  "top_p": 0.95,
+  "transformers_version": "4.51.3"
+}

model-00001-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18dbe98abe48f2c7d84cb28a2b97101b72634a2888e3d6eb7a8d6fbf5281b3ba
+size 4995884768

model-00002-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d0d18aabfc7cce4d182027358d7ffd4504ac6278c9c9ca12bb60dd8975ff86e
+size 4995052496

model-00003-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:673d355712864dbd1f27314c0360f0a963e0ee9452ea36c8ae42bee4a4990756
+size 4999071368

model-00004-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7b4f4daa2b95ce73e353a8fef1f049bc07c75125bbc6fa601b554440adadd81
+size 4994943096

model-00005-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:effea69cc5740f7ea8b94d7b4af9311c483a4b5bdd59b5eeb0a36cd47f7e6ae1
+size 4994943232

model-00006-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:683ef342bcc6c2d72c94e35d35b384f9821ef27afb9d78573279fd287aa9e385
+size 4994943456

model-00007-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e4837be87f1a2876a0012f4da8b380b06c4a683b53974195ff24b2b7411bcad
+size 4999180624

model-00008-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5fe73fa25a336e6be2707c63474d96b0add20015a4317dde154953ffcc343d2
+size 4994943096

model-00009-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:165cdcb1f16f14ef6a83851960d46375dbaabf7c172336cf38456942c879303c
+size 4994943096

model-00010-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d55d0234e36694eb8575633f04612cbee1be0b055f5f41d8519975a7060eb353
+size 4994943720

model-00011-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08a97b781181af8e0c81d0837a559b13711480de19de251df87c61a2ab2f50f3
+size 4916044152

model-00012-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57b62c10bde8448a29f71160465ba9cf418545e124fa145ca9a6c855745cd42a
+size 5000091936

model-00013-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fe8b74198c04e16b7a0301385e2248815338eb0af42aca2efb7e4b100fa9ee5
+size 4994944296

model-00014-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81dc88daf114818e8ad9866ef85c403869b7b4391af2f92bc5092344e4a3df55
+size 4994944392

model-00015-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28a0848194500e9f8daff3e3c3b20ec8b52cea81b8b7869e7a2daad7b0f2b5f9
+size 4994944648

model-00016-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce8407bd0d3ac2397a66e75f86034e7789cb547846b2bea484d293f1222041de
+size 4999181952

model-00017-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37f9bbf0b681bdbb51f09045a4894cc4c91ff86aca6a040a03d959a1b04f86c9
+size 4994944288

model-00018-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c67808e128166d99dd2285e6454668bdbb205a9d9efa88bcc51a4187fce22282
+size 4994944296

model-00019-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4d41bad246c7fca9ea4b052e3200c84642f20e855e3d1caf78f5dd7ec9f7cc8
+size 4994944536

model-00020-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:062665046cb9cb0e49f0ca7a194cd0b11a22e57e2604e6d75231705bb274207c
+size 4767862480

model-00021-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a109448c141bbca3c63ec74444b6bdf07ac997acd9e3d25d935403f6ea57f576
+size 5000091936

model-00022-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ba1783ead4f05432085a1de14f0fb7c19de024f8de671fd38881fc046ade127
+size 4994944296

model-00023-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5219488f545c644e71eb3d1eb75032c34dd89601a6249f29533336a0f1a6e4c0
+size 4994944392

model-00024-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9ee834cc694739ddb1b118a3658c600ebc81f0f3b8169f33940bb4501dc5742
+size 4994944648

model-00025-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b96f16d81ccbdbbc7ab261be03bc07070c2bc00e27a6e47cf7f36b87463017e
+size 4999181952

model-00026-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcf31ec6ae326831ccfd684b99e4642193606baf2c116f3903b57c81e1e28fa8
+size 4994944288

model-00027-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:059afaaf6c34ee1d1826fe85f859d4a3a86c404f274eaf9e3e871e168c405fc6
+size 4994944296

model-00028-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4d2b6363a3be0f08947d8f83d71f5b9985ef67acb231db25d1b0d96216c15b1
+size 4994944536

model-00029-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0ae064e782c09e2656598ffd40d71dca996e8fdc6c9a06397e976b91eb83914
+size 4767862480

model-00030-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd6637572be2f16e282a1681e30c3d35ebe60d29f7a43e3efaa69be2cf6e8f6e
+size 5000091936

model-00031-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3601bdcbc7cf88fd65faf66f82b096e58b8efe5753278e914a253e7c9d65af17
+size 4994944296

model-00032-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:282a65080530b3414164124adea2fc874343eec6393010780818a5a581677a86
+size 4994944392

model-00033-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:509b077f714559ea4084aabd88bfaecf3dfd0f93250b9f3b836512b0b29dafb4
+size 4994944648

model-00034-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b564d29b021df8fc56a990de50a0d2c7f0354636ed1ce39eb9109bb89a9962b
+size 4999181952

model-00035-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7a6772db37a06a7033f8dd902a0873eaccb897e4332b72bda489ddaed49ed5e
+size 4994944288

model-00036-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b087e16feb88f56ec6e75e2d2d6d5a40e2300754a7601b23cd4adb6b61ddebb
+size 4994944296

model-00037-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:503e5e5d916120f954937fa31f8a8868e07fb2ae3c0d3851f80add2c864a5f2b
+size 4994944536

model-00038-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cd876c9d8b73a45da3e38fc2245da325d8bd8087c16bde87e972b16ad87d4cb
+size 4767862480

model-00039-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68299e7dd1be2f23bde4b3edd8035f1eb0f27e118f245ebea5c30f7c0d0ba748
+size 5000091936

model-00040-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b18d545c348f00cf557de60f50a77f421c0375c7bcf4efec8ae707df2bc9d294
+size 4994944296

model-00041-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc2ccd0ea0b2ba73074674c8c66b994b94711f3f89452dfc5f527c7b5ec4db84
+size 4994944392

model-00042-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49a1d0d5712e4883ec88de234f83bd555c028052b525c4d75aa7cdf164fe252e
+size 4994944648

model-00043-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3777ce084609b712999c04974765608ed8fc53d53dbb4b9a32aa4edc5e31e759
+size 4999181952

model-00044-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c5aace2e6d240cd46e71c64a2b1a6c54a910428d7aacddf290ee4f2283ab36e
+size 4994944288

model-00045-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22300a5cbe541012f4eb354ae2998cb49cb8fdaa2b5471d1c48dd97d297ca56e
+size 4994944296

model-00046-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fb7f3ecd38ece6ead75a735b22ad44d9f84fd461cdfbf3f822cbcef70726f66
+size 4994944536

model-00047-of-00076.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e9cea92f2e38b2f5f953838410058530cdb5213efcf2d5413cba6bce067db7d
+size 4767862480