| { | |
| "architectures": [ | |
| "FLASHTransformerForPretrained" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "dnaflash.FLASHTransformerConfig", | |
| "AutoModel": "dnaflash.FLASHTransformerForPretrained", | |
| "AutoModelForSequenceClassification": "dnaflash.FLASHTransformerForSequenceClassification" | |
| }, | |
| "attn_dropout": 0.0, | |
| "causal": false, | |
| "expansion_factor": 2.0, | |
| "group_size": 256, | |
| "hidden_size": 1024, | |
| "laplace_attn_fn": false, | |
| "model_type": "flash_transformer", | |
| "norm_type": "scalenorm", | |
| "num_layers": 36, | |
| "query_key_dim": 128, | |
| "reduce_group_non_causal_attn": true, | |
| "shift_tokens": true, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.39.3", | |
| "vocab_size": 4096 | |
| } | |