| |
| |
|
|
| from transformers import PreTrainedConfig |
|
|
|
|
| class TrorYongASRConfig(PreTrainedConfig): |
| model_type = "troryongasr" |
|
|
| def __init__( |
| self, |
| vocab_size: int = 48005, |
| n_mels: int = 80, |
| n_audio_ctx: int = 1500, |
| n_text_ctx: int = 1024, |
| n_embed: int = 384, |
| n_head: int = 6, |
| n_layer: int = 4, |
| dropout: float= 0.0, |
| bias: bool = True, |
| pad_id: int = 0, |
| eot_id: int = 2, |
| tie_word_embeddings: bool = True, |
| **kwargs, |
| ): |
| self.vocab_size = vocab_size |
| self.n_mels = n_mels |
| self.n_audio_ctx = n_audio_ctx |
| self.n_text_ctx = n_text_ctx |
| self.n_embed = n_embed |
| self.n_head = n_head |
| self.n_layer = n_layer |
| self.dropout = dropout |
| self.bias = bias |
| self.pad_id = pad_id |
| self.eot_id = eot_id |
| self.tie_word_embeddings = tie_word_embeddings |
| super().__init__(**kwargs) |
|
|
