| { | |
| "architectures": [ | |
| "FHN_T4Max_150M_SingleHead_GAU" | |
| ], | |
| "model_type": "fhn_concept_gau", | |
| "hidden_size": 512, | |
| "expansion_dim": 1024, | |
| "attn_dim": 192, | |
| "num_attention_heads": 1, | |
| "num_hidden_layers": 24, | |
| "vocab_size": 3072, | |
| "concept_size": 2048, | |
| "max_position_embeddings": 512, | |
| "logit_soft_capping": 20.0, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "float16", | |
| "transformers_version": "4.44.0" | |
| } |