| { | |
| "architectures": [ | |
| "DeepseekForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "auto_map": { | |
| "AutoConfig": "configuration_deepseek.DeepseekConfig", | |
| "AutoModel": "modelling_deepseek.DeepseekModel", | |
| "AutoModelForCausalLM": "modelling_deepseek.DeepseekForCausalLM" | |
| }, | |
| "aux_loss_alpha": 0.001, | |
| "bos_token_id": 1, | |
| "eos_token_id": 128001, | |
| "first_k_dense_replace": 1, | |
| "head_dim": 128, | |
| "hidden_act": "silu", | |
| "hidden_size": 2048, | |
| "initializer_range": 0.006, | |
| "intermediate_size": 14336, | |
| "max_position_embeddings": 131072, | |
| "mlp_bias": false, | |
| "model_type": "deepseek", | |
| "moe_implementation": "eager", | |
| "moe_intermediate_size": 1792, | |
| "moe_layer_freq": 1, | |
| "n_routed_experts": 64, | |
| "n_shared_experts": 2, | |
| "norm_topk_prob": false, | |
| "num_attention_heads": 16, | |
| "num_experts_per_tok": 6, | |
| "num_hidden_layers": 28, | |
| "num_key_value_heads": 8, | |
| "pad_token_id": 1, | |
| "pretraining_tp": 1, | |
| "rms_norm_eps": 1e-05, | |
| "rope_scaling": null, | |
| "rope_theta": 1400000, | |
| "scoring_func": "softmax", | |
| "seq_aux": true, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.47.0", | |
| "use_cache": true, | |
| "vocab_size": 128256 | |
| } | |