Send response.
(self, out: InferOutput)
| 196 | logger.debug(f'Response: num_outputs={len(outputs)}.') |
| 197 | |
| 198 | def _send_resp(self, out: InferOutput): |
| 199 | """Send response.""" |
| 200 | logprobs = None if out.resp.data is None else out.resp.data.get('logprobs', None) |
| 201 | if out.finish: |
| 202 | resp_type = ResponseType.FINISH |
| 203 | elif out.resp.is_done: |
| 204 | resp_type = out.resp.type |
| 205 | else: |
| 206 | resp_type = ResponseType.SUCCESS |
| 207 | response_reqs(self.req_manager, |
| 208 | out.resp, |
| 209 | resp_type, |
| 210 | data=dict(token_ids=out.token_ids, |
| 211 | logits=out.logits, |
| 212 | cache_block_ids=out.cache_block_ids, |
| 213 | req_metrics=out.req_metrics, |
| 214 | routed_experts=out.routed_experts, |
| 215 | logprobs=logprobs)) |
| 216 | |
| 217 | @staticmethod |
| 218 | def _update_logprobs(step_outputs: list[InferOutput]): |