| { | |
| "architectures": [ | |
| "BitNetForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 128000, | |
| "eos_token_id": 128001, | |
| "hidden_act": "relu2", | |
| "hidden_size": 512, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 1024, | |
| "max_position_embeddings": 2048, | |
| "model_type": "bitnet", | |
| "num_attention_heads": 8, | |
| "num_hidden_layers": 6, | |
| "num_key_value_heads": 4, | |
| "quantization_config": { | |
| "linear_class": "autobitlinear", | |
| "quant_method": "bitnet", | |
| "quantization_mode": "offline" | |
| }, | |
| "rms_norm_eps": 1e-05, | |
| "rope_theta": 500000.0, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.52.0", | |
| "use_cache": true, | |
| "vocab_size": 128256 | |
| } | |