| { | |
| "model_type": "cosmicfish", | |
| "architectures": [ | |
| "CosmicFish" | |
| ], | |
| "vocab_size": 50257, | |
| "n_embd": 640, | |
| "n_layer": 10, | |
| "n_head": 16, | |
| "block_size": 512, | |
| "bias": true, | |
| "dropout": 0.1, | |
| "eps": 1e-06, | |
| "use_rotary": true, | |
| "use_swiglu": true, | |
| "use_gqa": true, | |
| "use_qk_norm": false, | |
| "n_query_groups": 4, | |
| "torch_dtype": "float16", | |
| "transformers_version": "4.36.0", | |
| "use_cache": true, | |
| "pad_token_id": 50256, | |
| "bos_token_id": 50256, | |
| "eos_token_id": 50256 | |
| } | |