Modified to support custom all reduce by default (#3538)

This commit is contained in:
zhink
2025-08-22 16:59:05 +08:00
committed by GitHub
parent 27666ee586
commit df7c31012b
15 changed files with 18 additions and 30 deletions

View File

@@ -35,7 +35,7 @@
| ```reasoning_parser``` | `str` | 指定要使用的推理解析器,以便从模型输出中提取推理内容 |
| ```use_cudagraph``` | `bool` | 是否使用cuda graph默认False。开启前建议仔细阅读 [graph_optimization.md](./features/graph_optimization.md),在多卡场景需要同时开启 Custom all-reduce。 |
| ```graph_optimization_config``` | `dict[str]` | 可以配置计算图优化相关的参数,默认值为'{"use_cudagraph":false, "graph_opt_level":0, "cudagraph_capture_sizes": null }',详细说明参考 [graph_optimization.md](./features/graph_optimization.md)|
| ```enable_custom_all_reduce``` | `bool` | 开启Custom all-reduce默认False |
| ```disable_custom_all_reduce``` | `bool` | 关闭Custom all-reduce默认False |
| ```splitwise_role``` | `str` | 是否开启splitwise推理默认值mixed 支持参数为["mixed", "decode", "prefill"] |
| ```innode_prefill_ports``` | `str` | prefill 实例内部引擎启动端口 仅单机PD分离需要默认值None |
| ```guided_decoding_backend``` | `str` | 指定要使用的guided decoding后端支持 `auto`、`xgrammar`、`off`, 默认为 `off` |