| { | |
| "architectures": [ | |
| "DeepseekV3ForCausalLM" | |
| ], | |
| "attention_backend": "flash_attention_2", | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 50256, | |
| "classifier_dropout": null, | |
| "dtype": "float32", | |
| "eos_token_id": 50256, | |
| "ep_size": 1, | |
| "first_k_dense_replace": 1, | |
| "head_dim": 32, | |
| "hidden_act": "silu", | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 256, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 1024, | |
| "kv_lora_rank": 64, | |
| "max_position_embeddings": 1024, | |
| "model_type": "deepseek_v3", | |
| "moe_intermediate_size": 128, | |
| "moe_layer_freq": 2, | |
| "n_group": 1, | |
| "n_routed_experts": 4, | |
| "n_shared_experts": 1, | |
| "norm_topk_prob": true, | |
| "num_attention_heads": 8, | |
| "num_experts_per_tok": 2, | |
| "num_hidden_layers": 6, | |
| "num_key_value_heads": 8, | |
| "num_nextn_predict_layers": 1, | |
| "pad_token_id": 50256, | |
| "pretraining_tp": 1, | |
| "q_lora_rank": 96, | |
| "qk_head_dim": 32, | |
| "qk_nope_head_dim": 0, | |
| "qk_rope_head_dim": 32, | |
| "rms_norm_eps": 1e-06, | |
| "rope_interleave": true, | |
| "rope_scaling": null, | |
| "rope_theta": 10000.0, | |
| "routed_scaling_factor": 1, | |
| "scoring_func": "softmax", | |
| "tie_word_embeddings": true, | |
| "topk_group": 1, | |
| "topk_method": "softmax_aux", | |
| "transformers_version": "4.56.0", | |
| "use_cache": false, | |
| "v_head_dim": 32, | |
| "vocab_size": 50257 | |
| } | |