Deviad's picture
Add files using upload-large-folder tool
399a4fb verified
{
"architectures": [
"DeepseekV4ForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 0,
"eos_token_id": 1,
"expert_dtype": "fp4",
"hc_eps": 1e-06,
"hc_mult": 4,
"hc_sinkhorn_iters": 20,
"head_dim": 512,
"hidden_act": "silu",
"hidden_size": 4096,
"index_head_dim": 128,
"index_n_heads": 64,
"index_topk": 512,
"initializer_range": 0.02,
"max_position_embeddings": 1048576,
"model_type": "deepseek_v4",
"moe_intermediate_size": 2048,
"n_routed_experts": 256,
"n_shared_experts": 1,
"norm_topk_prob": true,
"num_attention_heads": 64,
"num_experts_per_tok": 6,
"num_hidden_layers": 43,
"num_hash_layers": 3,
"num_key_value_heads": 1,
"num_nextn_predict_layers": 1,
"o_groups": 8,
"o_lora_rank": 1024,
"q_lora_rank": 1024,
"qk_rope_head_dim": 64,
"rms_norm_eps": 1e-06,
"rope_scaling": {
"beta_fast": 32,
"beta_slow": 1,
"factor": 16,
"original_max_position_embeddings": 65536,
"type": "yarn"
},
"rope_theta": 10000,
"routed_scaling_factor": 1.5,
"scoring_func": "sqrtsoftplus",
"sliding_window": 128,
"swiglu_limit": 10.0,
"tie_word_embeddings": false,
"topk_method": "noaux_tc",
"torch_dtype": "bfloat16",
"transformers_version": "4.57.1",
"use_cache": true,
"vocab_size": 129280,
"compress_rope_theta": 160000,
"compress_ratios": [
0,
0,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
128,
4,
0
],
"quantization": {
"group_size": 32,
"bits": 8,
"mode": "affine",
"model.layers.0.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.0.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.0.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.1.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.1.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.1.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.2.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.2.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.2.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.3.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.3.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.3.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.4.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.4.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.4.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.5.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.5.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.5.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.6.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.6.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.6.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.7.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.7.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.7.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.8.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.8.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.8.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.9.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.9.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.9.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.10.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.10.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.10.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.11.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.11.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.11.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.12.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.12.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.12.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.13.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.13.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.13.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.14.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.14.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.14.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.15.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.15.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.15.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.16.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.16.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.16.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.17.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.17.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.17.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.18.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.18.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.18.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.19.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.19.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.19.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.20.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.20.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.20.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.21.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.21.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.21.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.22.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.22.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.22.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.23.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.23.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.23.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.24.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.24.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.24.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.25.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.25.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.25.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.26.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.26.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.26.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.27.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.27.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.27.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.28.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.28.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.28.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.29.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.29.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.29.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.30.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.30.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.30.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.31.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.31.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.31.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.32.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.32.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.32.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.33.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.33.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.33.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.34.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.34.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.34.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.35.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.35.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.35.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.36.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.36.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.36.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.37.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.37.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.37.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.38.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.38.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.38.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.39.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.39.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.39.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.40.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.40.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.40.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.41.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.41.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.41.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.42.mlp.switch_mlp.gate_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.42.mlp.switch_mlp.down_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
},
"model.layers.42.mlp.switch_mlp.up_proj": {
"group_size": 32,
"bits": 4,
"mode": "affine"
}
},
"_name_or_path": "DSV4-Flash-MLX-Q4Q8",
"routed_expert_bits": 4,
"group_size": 32,
"mxtq_seed": 42,
"rope_parameters": {
"beta_fast": 32.0,
"beta_slow": 1.0,
"factor": 16.0,
"original_max_position_embeddings": 65536,
"rope_type": "yarn",
"rope_theta": 10000.0
}
}