(self)
| 133 | self.cache = T.full((bsize, *cache_shape), CACHE_FILL_VALUE, device=device, dtype=dtype).transpose(0, 1) |
| 134 | |
| 135 | def deinit_cache(self): |
| 136 | self.cache = [None] * self.cache_num_layers |
| 137 | |
| 138 | @T.no_grad() |
| 139 | def forward(self, data, next_tokens: Optional[Tuple[T.Tensor, T.Tensor]] = None, temps: Optional[Tuple[float, Tuple[float, float]]] = None): |
nothing calls this directly
no outgoing calls
no test coverage detected