| { |
| "architectures": [ |
| "FlintForCausalLM" |
| ], |
| "model_type": "flint", |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "bos_token_id": 150000, |
| "decoder_sparse_step": 1, |
| "eos_token_id": [ |
| 150001, |
| 150003, |
| 150005, |
| 150018 |
| ], |
| "head_dim": 32, |
| "hidden_act": "silu", |
| "hidden_size": 32, |
| "initializer_range": 0.02, |
| "intermediate_size": 128, |
| "layer_types": [ |
| "full_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "sliding_attention", |
| "full_attention", |
| "sliding_attention", |
| "sliding_attention" |
| ], |
| "max_position_embeddings": 4096, |
| "mlp_only_layers": [ |
| 0, |
| 1 |
| ], |
| "moe_intermediate_size": 32, |
| "norm_topk_prob": true, |
| "num_attention_heads": 1, |
| "num_experts": 64, |
| "num_experts_per_tok": 6, |
| "num_hidden_layers": 8, |
| "num_key_value_heads": 1, |
| "no_rope_layer_indices": [ |
| 0, |
| 1, |
| 5 |
| ], |
| "output_router_logits": false, |
| "pad_token_id": 150001, |
| "rms_norm_eps": 1e-06, |
| "rope_theta": 10000.0, |
| "router_aux_loss_coef": 0.0005, |
| "router_score_function": "sigmoid", |
| "shared_expert_intermediate_size": 64, |
| "sliding_window": 513, |
| "softmax_type": "learnable", |
| "tie_word_embeddings": false, |
| "torch_dtype": "bfloat16", |
| "use_cache": true, |
| "use_sliding_window": true, |
| "vocab_size": 150144, |
| "transformers_version": "5.12.1" |
| } |
|
|
