| from transformers import PretrainedConfig | |
| class TamilTinyStoriesConfig(PretrainedConfig): | |
| model_type = "tamil_tiny_stories" | |
| def __init__( | |
| self, | |
| vocab_size=0, | |
| original_vocab_size=None, | |
| block_size=128, | |
| n_embd=128, | |
| n_head=4, | |
| n_layer=4, | |
| dropout=0.0, | |
| bos_token_id=None, | |
| eos_token_id=None, | |
| pad_token_id=None, | |
| unk_token_id=None, | |
| use_cache=False, | |
| **kwargs, | |
| ): | |
| self.vocab_size = vocab_size | |
| self.original_vocab_size = original_vocab_size if original_vocab_size is not None else vocab_size | |
| self.block_size = block_size | |
| self.n_embd = n_embd | |
| self.n_head = n_head | |
| self.n_layer = n_layer | |
| self.dropout = dropout | |
| self.hidden_size = n_embd | |
| self.num_attention_heads = n_head | |
| self.num_hidden_layers = n_layer | |
| self.max_position_embeddings = block_size | |
| self.use_cache = use_cache | |
| self.is_decoder = True | |
| super().__init__( | |
| bos_token_id=bos_token_id, | |
| eos_token_id=eos_token_id, | |
| pad_token_id=pad_token_id, | |
| unk_token_id=unk_token_id, | |
| **kwargs, | |
| ) | |
| TamilTinyStoriesConfig.register_for_auto_class() |