| { | |
| "_attn_implementation_autoset": true, | |
| "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, | |
| "dtype": "bfloat16", | |
| "eos_token_id": 1, | |
| "ep_size": 1, | |
| "first_k_dense_replace": 1, | |
| "head_dim": 64, | |
| "hidden_act": "silu", | |
| "hidden_size": 2048, | |
| "initializer_range": 0.0125, | |
| "intermediate_size": 10944, | |
| "kv_lora_rank": 512, | |
| "max_position_embeddings": 8192, | |
| "model_type": "deepseek_v3", | |
| "moe_intermediate_size": 1408, | |
| "moe_layer_freq": 1, | |
| "n_group": 8, | |
| "n_routed_experts": 64, | |
| "n_shared_experts": 2, | |
| "norm_topk_prob": true, | |
| "num_attention_heads": 16, | |
| "num_experts_per_tok": 8, | |
| "num_hidden_layers": 16, | |
| "num_key_value_heads": 16, | |
| "num_nextn_predict_layers": 1, | |
| "pretraining_tp": 1, | |
| "q_lora_rank": 1024, | |
| "qk_head_dim": 192, | |
| "qk_nope_head_dim": 128, | |
| "qk_rope_head_dim": 64, | |
| "rms_norm_eps": 1e-06, | |
| "rope_interleave": true, | |
| "rope_scaling": { | |
| "beta_fast": 32.0, | |
| "beta_slow": 1.0, | |
| "factor": 40.0, | |
| "mscale": 1.0, | |
| "mscale_all_dim": 1.0, | |
| "original_max_position_embeddings": 4096, | |
| "rope_type": "yarn", | |
| "type": "yarn" | |
| }, | |
| "rope_theta": 10000, | |
| "routed_scaling_factor": 1.0, | |
| "scoring_func": "sigmoid", | |
| "seq_aux": true, | |
| "tie_word_embeddings": false, | |
| "topk_group": 4, | |
| "topk_method": "noaux_tc", | |
| "transformers_version": "4.57.6", | |
| "use_cache": true, | |
| "v_head_dim": 128, | |
| "vocab_size": 129280, | |
| "quantization_config": { | |
| "config_groups": { | |
| "group_0": { | |
| "input_activations": { | |
| "dynamic": false, | |
| "num_bits": 4, | |
| "type": "float", | |
| "group_size": 16 | |
| }, | |
| "weights": { | |
| "dynamic": false, | |
| "num_bits": 4, | |
| "type": "float", | |
| "group_size": 16 | |
| }, | |
| "targets": [ | |
| "Linear" | |
| ] | |
| } | |
| }, | |
| "ignore": [ | |
| "lm_head", | |
| "model.layers.0.self_attn.kv_a_proj_with_mqa", | |
| "model.layers.0.self_attn.kv_b_proj", | |
| "model.layers.0.self_attn.q_a_proj", | |
| "model.layers.0.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.q_a_proj", | |
| "model.layers.1.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.q_a_proj", | |
| "model.layers.10.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.q_a_proj", | |
| "model.layers.11.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.q_a_proj", | |
| "model.layers.12.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.q_a_proj", | |
| "model.layers.13.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.q_a_proj", | |
| "model.layers.14.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.q_a_proj", | |
| "model.layers.15.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.q_a_proj", | |
| "model.layers.2.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.q_a_proj", | |
| "model.layers.3.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.q_a_proj", | |
| "model.layers.4.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.q_a_proj", | |
| "model.layers.5.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.q_a_proj", | |
| "model.layers.6.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.q_a_proj", | |
| "model.layers.7.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.q_a_proj", | |
| "model.layers.8.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.q_a_proj", | |
| "model.layers.9.self_attn.q_b_proj" | |
| ], | |
| "quant_algo": "NVFP4", | |
| "producer": { | |
| "name": "modelopt", | |
| "version": "0.41.0" | |
| }, | |
| "quant_method": "modelopt" | |
| } | |
| } |