| { | |
| "architectures": [ | |
| "RetNetModelWithLMHead" | |
| ], | |
| "chunk_size": 512, | |
| "eos_token_id": 0, | |
| "ffn_proj_size": 4096, | |
| "forward_impl": "parallel", | |
| "hidden_size": 2048, | |
| "initializer_range": 0.02, | |
| "is_decoder": true, | |
| "model_type": "retnet", | |
| "num_heads": 8, | |
| "num_layers": 24, | |
| "output_retentions": false, | |
| "pad_token_id": 0, | |
| "qk_dim": 2048, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.33.1", | |
| "unk_token_id": 0, | |
| "use_bias_in_mlp": true, | |
| "use_bias_in_msr": false, | |
| "use_bias_in_msr_out": false, | |
| "use_cache": true, | |
| "use_default_gamma": false, | |
| "v_dim": 4096, | |
| "vocab_size": 64000 | |
| } | |