| { |
| "architectures": [ |
| "AfmoeForCausalLM" |
| ], |
| "attention_dropout": 0.0, |
| "auto_map": { |
| "AutoConfig": "configuration_afmoe.AfmoeConfig", |
| "AutoModel": "modeling_afmoe.AfmoeModel", |
| "AutoModelForCausalLM": "modeling_afmoe.AfmoeForCausalLM" |
| }, |
| "dtype": "bfloat16", |
| "eos_token_id": 3, |
| "global_attn_every_n_layers": 4, |
| "head_dim": 128, |
| "hidden_act": "silu", |
| "hidden_size": 3072, |
| "initializer_range": 0.02, |
| "intermediate_size": 12288, |
| "layer_types": [ |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention" |
| ], |
| "load_balance_coeff": 5e-05, |
| "max_position_embeddings": 262144, |
| "model_type": "afmoe", |
| "moe_intermediate_size": 3072, |
| "mup_enabled": true, |
| "n_group": 1, |
| "num_attention_heads": 48, |
| "num_dense_layers": 6, |
| "num_expert_groups": 1, |
| "num_experts": 256, |
| "num_experts_per_tok": 4, |
| "num_hidden_layers": 60, |
| "num_key_value_heads": 8, |
| "num_limited_groups": 1, |
| "num_shared_experts": 1, |
| "quantization": { |
| "group_size": 64, |
| "bits": 5, |
| "mode": "affine", |
| "model.layers.6.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.7.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.8.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.9.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.10.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.11.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.12.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.13.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.14.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.15.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.16.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.17.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.18.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.19.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.20.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.21.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.22.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.23.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.24.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.25.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.26.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.27.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.28.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.29.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.30.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.31.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.32.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.33.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.34.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.35.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.36.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.37.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.38.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.39.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.40.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.41.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.42.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.43.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.44.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.45.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.46.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.47.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.48.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.49.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.50.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.51.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.52.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.53.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.54.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.55.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.56.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.57.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.58.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.59.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| } |
| }, |
| "quantization_config": { |
| "group_size": 64, |
| "bits": 5, |
| "mode": "affine", |
| "model.layers.6.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.7.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.8.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.9.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.10.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.11.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.12.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.13.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.14.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.15.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.16.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.17.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.18.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.19.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.20.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.21.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.22.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.23.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.24.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.25.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.26.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.27.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.28.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.29.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.30.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.31.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.32.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.33.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.34.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.35.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.36.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.37.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.38.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.39.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.40.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.41.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.42.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.43.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.44.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.45.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.46.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.47.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.48.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.49.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.50.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.51.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.52.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.53.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.54.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.55.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.56.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.57.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.58.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| }, |
| "model.layers.59.mlp.router.gate": { |
| "group_size": 64, |
| "bits": 8 |
| } |
| }, |
| "rms_norm_eps": 1e-05, |
| "rope_scaling": null, |
| "rope_theta": 10000, |
| "route_norm": true, |
| "route_scale": 2.448, |
| "score_func": "sigmoid", |
| "sliding_window": 4096, |
| "tie_word_embeddings": false, |
| "topk_group": 1, |
| "transformers_version": "4.57.1", |
| "use_cache": true, |
| "use_grouped_mm": true, |
| "vocab_size": 200192 |
| } |
