[QUESTION]你好，flux/test/python/moe_ag_scatter/test_moe_ag.py这个需要在8gpu的时候设置tp size=2需要怎么改？

[jinchen89]

我看flux/test/python/moe_gather_rs/test_moe_gather_rs.py里面可以随意设置tp size？我理解tp size可以和gpu数量不相等的，请不吝赐教，感谢！

[ZSL98]

In test_moe_ag.py, the tp size you want is actually ffn_tp_size

[jinchen89]

所以TP_GROUP.size()指的是物理gpu的数量？

[ZSL98]

In the context of sequence parallelism, the input is partitioned into shards, with the number of shards being equal to TP_GROUP.size(), as is the situation in our example. In this scenario, the TP_GROUP.size() is equivalent to the number of devices.

[LevinSx]

Just change the code in launch.sh nproc_per_node=2