| { |
| "architectures": [ |
| "MapleForCausalLM" |
| ], |
| "attention_dropout": 0.0, |
| "bos_token_id": 151643, |
| "dtype": "bfloat16", |
| "embedding_dropout": 0.0, |
| "eos_token_id": 151645, |
| "first_k_dense_replace": 0, |
| "flash_head": { |
| "bits": 4, |
| "cluster_size": 32, |
| "force_tokens": [ |
| 151645, |
| 151668, |
| 151643 |
| ], |
| "group_size": 64, |
| "head_bits": 4, |
| "head_group_size": 64, |
| "n_clusters": 4748, |
| "n_probes": 512, |
| "scaled_centroids": true |
| }, |
| "hc": false, |
| "head_dim": 128, |
| "headwise_gate_swa": false, |
| "hidden_act": "silu", |
| "hidden_size": 2048, |
| "initializer_range": 0.02, |
| "intermediate_size": 5120, |
| "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" |
| ], |
| "max_position_embeddings": 128000, |
| "max_window_layers": 24, |
| "model_file": "maple.py", |
| "model_type": "maple", |
| "moe_intermediate_size": 512, |
| "moe_router_enable_expert_bias": false, |
| "moe_shared_expert_intermediate_size": 512, |
| "mtp_loss_scaling_factor": 0, |
| "nope_on_global_attention": false, |
| "norm_topk_prob": true, |
| "num_attention_heads": 16, |
| "num_experts": 256, |
| "num_experts_per_tok": 8, |
| "num_hidden_layers": 24, |
| "num_key_value_heads": 4, |
| "num_nextn_predict_layers": 0, |
| "num_shared_experts": 0, |
| "output_dropout": 0.0, |
| "output_router_logits": false, |
| "pad_token_id": null, |
| "partial_rotary_factor": 0.5, |
| "preaffine": false, |
| "quantization": { |
| "bits": 2, |
| "group_size": 128, |
| "lm_head": { |
| "bits": 4, |
| "group_size": 64 |
| }, |
| "mode": "affine", |
| "model.word_embeddings": { |
| "bits": 4, |
| "group_size": 64 |
| } |
| }, |
| "quantization_config": { |
| "bits": 2, |
| "group_size": 128, |
| "lm_head": { |
| "bits": 4, |
| "group_size": 64 |
| }, |
| "mode": "affine", |
| "model.word_embeddings": { |
| "bits": 4, |
| "group_size": 64 |
| } |
| }, |
| "rms_norm_eps": 1e-06, |
| "rope_scaling": null, |
| "rope_theta": 10000, |
| "router_dtype": "fp32", |
| "sliding_window": 512, |
| "tie_word_embeddings": false, |
| "transformers_version": "4.57.1", |
| "use_bias": false, |
| "use_cache": true, |
| "use_qk_norm": true, |
| "use_qkv_bias": false, |
| "use_rmsnorm": true, |
| "vocab_size": 151936 |
| } |
