{ | |
"dim": 320, | |
"n_layers": 12, | |
"n_heads": 20, | |
"vocab_size": 2362, | |
"dim_ffn": 960, | |
"norm_eps": 1e-05, | |
"multiple_of": 64, | |
"padding_idx": 2, | |
"max_seq_len": 512, | |
"initializer_range": 0.02 | |
} |
{ | |
"dim": 320, | |
"n_layers": 12, | |
"n_heads": 20, | |
"vocab_size": 2362, | |
"dim_ffn": 960, | |
"norm_eps": 1e-05, | |
"multiple_of": 64, | |
"padding_idx": 2, | |
"max_seq_len": 512, | |
"initializer_range": 0.02 | |
} |