fix gid (#4038)

gzy19990617 · web-flow · commit b41988f4bcd5 · 2025-09-16T20:56:36.000+08:00
diff --git a/fastdeploy/config.py b/fastdeploy/config.py
@@ -349,12 +349,14 @@ def set_tp_group(self):
                 (self.data_parallel_rank + 1) * self.tensor_parallel_size,
             )
         )
+        dist.collective._set_custom_gid(None)
         # same ep group id
         dist.collective._set_custom_gid(self.data_parallel_size + tp_gid_offset)
         self.ep_group = dist.new_group(range(self.expert_parallel_size))
         logger.info(
             f"data_parallel_size: {self.data_parallel_size}, tensor_parallel_size: {self.tensor_parallel_size}, expert_parallel_size: {self.expert_parallel_size}, data_parallel_rank: {self.data_parallel_rank}, tensor_parallel_rank: {self.tensor_parallel_rank}, expert_parallel_rank: {self.expert_parallel_rank}, tp_group: {self.tp_group}."
         )
+        dist.collective._set_custom_gid(None)
 
     def print(self):
         """

Original file line number	Diff line number	Diff line change
`@@ -349,12 +349,14 @@ def set_tp_group(self):`
`349`	`349`	`(self.data_parallel_rank + 1) * self.tensor_parallel_size,`
`350`	`350`	`)`
`351`	`351`	`)`
	`352`	`+ dist.collective._set_custom_gid(None)`
`352`	`353`	`# same ep group id`
`353`	`354`	`dist.collective._set_custom_gid(self.data_parallel_size + tp_gid_offset)`
`354`	`355`	`self.ep_group = dist.new_group(range(self.expert_parallel_size))`
`355`	`356`	`logger.info(`
`356`	`357`	`f"data_parallel_size: {self.data_parallel_size}, tensor_parallel_size: {self.tensor_parallel_size}, expert_parallel_size: {self.expert_parallel_size}, data_parallel_rank: {self.data_parallel_rank}, tensor_parallel_rank: {self.tensor_parallel_rank}, expert_parallel_rank: {self.expert_parallel_rank}, tp_group: {self.tp_group}."`
`357`	`358`	`)`
	`359`	`+ dist.collective._set_custom_gid(None)`
`358`	`360`
`359`	`361`	`def print(self):`
`360`	`362`	`"""`