| { | |
| "architectures": [ | |
| "GptOssForCausalLM" | |
| ], | |
| "attention_bias": true, | |
| "attention_dropout": 0.0, | |
| "eos_token_id": 200002, | |
| "experts_per_token": 4, | |
| "head_dim": 64, | |
| "hidden_act": "silu", | |
| "hidden_size": 2880, | |
| "initial_context_length": 4096, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 2880, | |
| "layer_types": [ | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "full_attention" | |
| ], | |
| "max_position_embeddings": 131072, | |
| "model_type": "gpt_oss", | |
| "num_attention_heads": 64, | |
| "num_experts_per_tok": 4, | |
| "num_hidden_layers": 36, | |
| "num_key_value_heads": 8, | |
| "num_local_experts": 128, | |
| "output_router_logits": false, | |
| "pad_token_id": 199999, | |
| "quantization": { | |
| "group_size": 32, | |
| "bits": 4, | |
| "mode": "mxfp4", | |
| "model.embed_tokens": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.0.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.0.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.1.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.2.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.3.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.6.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.9.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.12.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.12.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.12.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.12.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.12.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.15.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.15.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.15.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.15.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.15.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.18.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.18.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.18.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.18.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.18.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.21.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.21.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.21.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.21.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.21.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.24.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.24.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.24.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.24.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.24.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.27.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.27.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.27.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.27.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.27.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.30.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.30.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.30.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.30.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.30.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.31.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.31.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.31.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.31.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.31.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.32.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.32.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.32.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.32.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.32.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.33.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.33.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.33.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.33.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.33.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.34.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.34.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.34.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.34.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.34.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.35.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.35.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.35.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.35.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.35.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "lm_head": { | |
| "group_size": 64, | |
| "bits": 6 | |
| } | |
| }, | |
| "quantization_config": { | |
| "group_size": 32, | |
| "bits": 4, | |
| "mode": "mxfp4", | |
| "model.embed_tokens": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.0.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.0.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.1.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.2.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.3.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.4.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.5.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.6.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.7.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.8.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.9.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.10.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.11.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.12.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.12.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.12.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.12.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.12.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.13.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.14.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.15.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.15.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.15.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.15.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.15.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.16.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.17.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.18.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.18.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.18.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.18.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.18.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.19.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.20.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.21.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.21.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.21.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.21.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.21.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.22.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.23.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.24.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.24.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.24.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.24.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.24.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.25.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.26.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.27.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.27.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.27.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.27.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.27.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.28.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.29.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.30.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.30.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.30.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.30.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.30.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.31.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.31.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.31.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.31.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.31.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.32.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.32.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.32.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.32.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.32.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.33.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.33.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.33.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.33.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.33.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.34.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.34.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.34.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.34.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.34.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.35.self_attn.q_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.35.self_attn.k_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.35.self_attn.v_proj": { | |
| "group_size": 32, | |
| "bits": 6 | |
| }, | |
| "model.layers.35.self_attn.o_proj": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "model.layers.35.mlp.router": { | |
| "group_size": 32, | |
| "bits": 4 | |
| }, | |
| "lm_head": { | |
| "group_size": 64, | |
| "bits": 6 | |
| } | |
| }, | |
| "rms_norm_eps": 1e-05, | |
| "rope_scaling": { | |
| "beta_fast": 32.0, | |
| "beta_slow": 1.0, | |
| "factor": 32.0, | |
| "original_max_position_embeddings": 4096, | |
| "rope_type": "yarn", | |
| "truncate": false | |
| }, | |
| "rope_theta": 150000, | |
| "router_aux_loss_coef": 0.9, | |
| "sliding_window": 128, | |
| "swiglu_limit": 7.0, | |
| "tie_word_embeddings": false, | |
| "transformers_version": "4.55.0.dev0", | |
| "use_cache": true, | |
| "vocab_size": 201088 | |
| } |