(self, config: CohereConfig)
| 152 | config_class = CohereConfig |
| 153 | |
| 154 | def __init__(self, config: CohereConfig): |
| 155 | transformer = CohereModel(config) |
| 156 | vocab_size_padded = pad_vocab_size(config.vocab_size, |
| 157 | config.mapping.tp_size) |
| 158 | if config.mapping.is_last_pp_rank(): |
| 159 | lm_head = ColumnLinear(config.hidden_size, |
| 160 | vocab_size_padded, |
| 161 | bias=False, |
| 162 | dtype=config.dtype, |
| 163 | tp_group=config.mapping.tp_group, |
| 164 | tp_size=config.mapping.tp_size, |
| 165 | gather_output=True) |
| 166 | else: |
| 167 | lm_head = None |
| 168 | self.quant_mode = config.quant_mode |
| 169 | self.mapping = config.mapping |
| 170 | super().__init__(config, transformer, lm_head) |
| 171 | |
| 172 | @classmethod |
| 173 | def from_hugging_face(cls, |
nothing calls this directly
no test coverage detected