| { | |
| "model_type": "bitnet", | |
| "architectures": [ | |
| "BitNetModel" | |
| ], | |
| "torch_dtype": "float16", | |
| "_name_or_path": "bitnet-model", | |
| "vocab_size": 128256, | |
| "hidden_size": 1024, | |
| "max_position_embeddings": 128, | |
| "num_hidden_layers": 12, | |
| "num_attention_heads": 8, | |
| "intermediate_size": 4096, | |
| "hidden_act": "silu", | |
| "hidden_dropout_prob": 0.1, | |
| "attention_probs_dropout_prob": 0.1, | |
| "initializer_range": 0.02, | |
| "layer_norm_eps": 1e-05, | |
| "use_cache": false, | |
| "tie_word_embeddings": false, | |
| "pad_token_id": 0, | |
| "bos_token_id": 1, | |
| "eos_token_id": 2, | |
| "use_layer_skipping": true, | |
| "skip_probability": 0.1, | |
| "min_layers_to_keep": 4, | |
| "use_early_exit": true, | |
| "early_exit_threshold": 0.95, | |
| "gradient_checkpointing": false, | |
| "max_grad_norm": 1.0, | |
| "learning_rate": 0.0006, | |
| "warmup_steps": 1000, | |
| "weight_decay": 0.1, | |
| "attention_bias": false, | |
| "attention_dropout": 0.1, | |
| "rms_norm_eps": 1e-05 | |
| } |