| { | |
| "activation_dropout": 0.0, | |
| "activation_function": "gelu", | |
| "architectures": [ | |
| "OPTModel" | |
| ], | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 2, | |
| "d_model": 768, | |
| "decoder_layernorm": true, | |
| "decoder_start_token_id": 2, | |
| "dropout": 0.1, | |
| "embed_dim": 768, | |
| "eos_token_id": 2, | |
| "ffn_dim": 3072, | |
| "forced_eos_token_id": 2, | |
| "init_std": 0.02, | |
| "layerdrop": 0.0, | |
| "max_position_embeddings": 2048, | |
| "model_type": "opt", | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 12, | |
| "output_projection": false, | |
| "pad_token_id": 1, | |
| "scale_embedding": false, | |
| "share_input_output_embed": true, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.19.0.dev0", | |
| "use_cache": false, | |
| "vocab_size": 50272 | |
| } | |