| from transformers import GPT2Config | |
| class CubeConfig(GPT2Config): | |
| model_type = "CubeLM" | |
| def __init__( | |
| self, | |
| vocab_size=16, | |
| bos_token_id=15, | |
| eos_token_id=15, | |
| pad_token_id=15, | |
| n_positions=40, | |
| n_embd=512, | |
| n_layer=8, | |
| n_head=8, | |
| **kwargs | |
| ): | |
| super().__init__(**kwargs) | |
| self.vocab_size = vocab_size | |
| self.n_positions = n_positions | |
| self.n_embd = n_embd | |
| self.n_layer = n_layer | |
| self.n_head = n_head | |
| self.bos_token_id = bos_token_id | |
| self.eos_token_id = eos_token_id | |
| self.pad_token_id = pad_token_id | |