end-2-end reborn model for mls-portuguese unsupervised phoneme recognition (iter2-stage1)
a7e4030
verified
{ | |
"architectures": [ | |
"RebornUASRModel" | |
], | |
"auto_map": { | |
"AutoConfig": "configuration_reborn.RebornUASRConfig", | |
"AutoModel": "modeling_reborn.RebornUASRModel" | |
}, | |
"discriminator_act_after_linear": false, | |
"discriminator_causal": true, | |
"discriminator_depth": 1, | |
"discriminator_dilation": 1, | |
"discriminator_dim": 256, | |
"discriminator_dropout": 0.0, | |
"discriminator_input_dim": 51, | |
"discriminator_kernel": 3, | |
"discriminator_linear_emb": false, | |
"discriminator_max_pool": false, | |
"discriminator_spectral_norm": false, | |
"discriminator_weight_norm": false, | |
"generator_bias": false, | |
"generator_bn_apply": false, | |
"generator_bn_init_weight": 30.0, | |
"generator_dilation": 1, | |
"generator_dropout": 0.0, | |
"generator_input_dim": 512, | |
"generator_kernel": 4, | |
"generator_output_dim": 51, | |
"generator_stride": 1, | |
"model_type": "reborn_uasr", | |
"phones": [ | |
"s", | |
"a", | |
"i", | |
"e", | |
"o", | |
"\u027e", | |
"\u00e6", | |
"\u014b", | |
"d", | |
"k", | |
"t", | |
"m", | |
"\u028a", | |
"r", | |
"l", | |
"e\u026a", | |
"p", | |
"y", | |
"v", | |
"u", | |
"n", | |
"z", | |
"b", | |
"\u0250\u0303", | |
"\u0259", | |
"t\u0283", | |
"f", | |
"\u0261", | |
"x", | |
"\u0250\u0303\u028a\u0303", | |
"\u0292", | |
"d\u0292", | |
"w", | |
"\u025b", | |
"j", | |
"\u0254", | |
"a\u028a", | |
"\u0283", | |
"\u0272", | |
"i\u028a", | |
"u\u0303", | |
"o\u0303", | |
"a\u026a", | |
"e\u028a", | |
"\u025b\u028a", | |
"\u028e", | |
"<SIL>" | |
], | |
"segmenter_dropout": 0.1, | |
"segmenter_hidden_dim": 512, | |
"segmenter_input_dim": 512, | |
"segmenter_kernel_size": 7, | |
"segmenter_type": "cnn", | |
"special_token_nums": 4, | |
"torch_dtype": "float32", | |
"transformers_version": "4.24.0" | |
} | |