{ | |
"d_model": 512, | |
"in_model": 1280, | |
"n_heads": 8, | |
"intermediate_size": 1280, | |
"n_layers": 8, | |
"norm_eps": 1e-12, | |
"expr_bin": 32, | |
"initializer_range": 0.02, | |
"attention_out_bias": true, | |
"norm_type": "layer_norm", | |
"attention_dropout": 0.0, | |
"mlp_dropout": 0.0 | |
} |