| { | |
| "architectures": [ | |
| "Transformer" | |
| ], | |
| "attn_dropout": 0.0, | |
| "d_hidden": 3072, | |
| "d_model": 1024, | |
| "embed_dropout": 0.0, | |
| "mlp": "blp", | |
| "mlp_bias": false, | |
| "mlp_dropout": 0.0, | |
| "modifier": null, | |
| "n_ctx": 256, | |
| "n_head": 8, | |
| "n_layer": 1, | |
| "n_vocab": 4096, | |
| "noise": null, | |
| "norm_bias": false, | |
| "normalization": null, | |
| "resid_dropout": 0.0, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.39.0" | |
| } | |