{ | |
"architectures": [ | |
"ArgonneModelParallel" | |
], | |
"block_size": 2048, | |
"dropout": 0.1, | |
"model_type": "argonne", | |
"n_embd": 1296, | |
"n_head": 12, | |
"n_layer": 12, | |
"torch_dtype": "float32", | |
"transformers_version": "4.47.0", | |
"vocab_size": 12000 | |
} | |
{ | |
"architectures": [ | |
"ArgonneModelParallel" | |
], | |
"block_size": 2048, | |
"dropout": 0.1, | |
"model_type": "argonne", | |
"n_embd": 1296, | |
"n_head": 12, | |
"n_layer": 12, | |
"torch_dtype": "float32", | |
"transformers_version": "4.47.0", | |
"vocab_size": 12000 | |
} | |