Regenerate: fused layout (1593 tensors, 31 shards), mxfp4 experts + int8 channel dense, appmana kernel block
d62259e 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, | |
| "quantization_config": { | |
| "quant_method": "dsv4_mxfp4_int8", | |
| "format": "mxfp4_int8_packed", | |
| "config_groups": { | |
| "experts_w4a16": { | |
| "weights": { | |
| "num_bits": 4, | |
| "type": "float", | |
| "format": "mxfp4", | |
| "symmetric": false, | |
| "group_size": 32, | |
| "strategy": "group", | |
| "scale_mode": "native_e8m0", | |
| "scale_dtype": "e8m0" | |
| }, | |
| "input_activations": { | |
| "num_bits": 16, | |
| "type": "float" | |
| }, | |
| "targets": [ | |
| "*.ffn.experts.routed_experts.w13_weight", | |
| "*.ffn.experts.routed_experts.w2_weight" | |
| ] | |
| }, | |
| "linears_w8a16": { | |
| "weights": { | |
| "num_bits": 8, | |
| "type": "int", | |
| "symmetric": true, | |
| "strategy": "channel" | |
| }, | |
| "input_activations": { | |
| "num_bits": 16, | |
| "type": "float" | |
| }, | |
| "targets": [ | |
| "*.attn.fused_wqa_wkv", | |
| "*.attn.wq_b", | |
| "*.attn.wo_a", | |
| "*.attn.wo_b", | |
| "*.attn.indexer.wq_b", | |
| "*.attn.indexer.compressor.wkv", | |
| "*.ffn.shared_experts.gate_up_proj", | |
| "*.ffn.shared_experts.down_proj" | |
| ] | |
| } | |
| }, | |
| "ignore": [ | |
| "embed", | |
| "head", | |
| "norm", | |
| "lm_head", | |
| "*norm.weight", | |
| "attn.attn_sink", | |
| "*.gate.*", | |
| "hc_*", | |
| "*.hc_attn_*", | |
| "*.hc_ffn_*" | |
| ] | |
| }, | |
| "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 | |
| ], | |
| "appmana": { | |
| "kernels": [ | |
| "flash_mla.flash_sparse_mla_decode", | |
| "flash_mla.triton_sparse_int8_mla_decode", | |
| "vllm.models.deepseek_v4.nvidia_sm86.triton_kernels.sparse_attention_triton" | |
| ], | |
| "cache_type": "fp8_ds_mla" | |
| } | |
| } |