| { | |
| "model_type": "residualnet", | |
| "architectures": [ | |
| "ResidualNetForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_residualnet.ResidualNetConfig", | |
| "AutoModel": "modeling_residualnet.ResidualNetModel", | |
| "AutoModelForCausalLM": "modeling_residualnet.ResidualNetForCausalLM" | |
| }, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": null, | |
| "embd_pdrop": 0.0, | |
| "eos_token_id": 151645, | |
| "hidden_act": "silu", | |
| "hidden_size": 128, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 64, | |
| "max_position_embeddings": 1024, | |
| "name": "residual-tiny", | |
| "num_attention_heads": 4, | |
| "num_hidden_layers": 4, | |
| "num_key_value_heads": 4, | |
| "original_max_position_embeddings": 1024, | |
| "pad_token_id": 151645, | |
| "resid_pdrop": 0.0, | |
| "rms_norm_eps": 1e-05, | |
| "rope_scaling": null, | |
| "rope_theta": 10000.0, | |
| "sliding_window": null, | |
| "tie_word_embeddings": true, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.48.2", | |
| "use_cache": true, | |
| "vocab_size": 151669 | |
| } | |