[Feat] support mixed ep #2969
Merged
[Feat] support mixed ep #2969
+137
−48
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Thanks for your contribution! |
| @@ -959,6 +959,11 @@ class at the server level, which is too granular for ModelRunner. | |||
| We plan to replace it with 'ModelForwardBatch'. | |||
| intermediate_tensors: | |||
| """ | |||
| is_decode_batch = paddle.to_tensor(not ((self.share_inputs["seq_lens_this_time"] | |||
| > 1).sum() > 0)) | |||
| paddle.distributed.broadcast(is_decode_batch, src=0) | |||
There was a problem hiding this comment.
这里不是broadcast的逻辑，而且src也不一定是0，这里是任意一张卡是False，其他卡必须是False，只有所有卡是True才能是True
Wanglongzhi2001
force-pushed
the
mixed_ep
branch
2 times, most recently
from
9a5badc to
34ebf2a
Compare
Wanglongzhi2001
force-pushed
the
mixed_ep
branch
from
34ebf2a to
c81a3f9
Compare
Wanglongzhi2001
commented
Jul 25, 2025
| num_max_dispatch_tokens_per_rank, | ||
| None, | ||
| num_experts, | ||
| ) |
There was a problem hiding this comment.
这段代码之前被删了，但是我发现3.0.1的paddle这里还是会报错，而Fastdeploy的使用文档里推荐用户使用的paddle版本还是3.0.1，所以从用户使用角度考虑先留着
There was a problem hiding this comment.
这里也需要适配develop 下一个PR这里都兼容一下吧
|
Batch 越大加速越明显吗 |
理论上是的，因为单机EP现在默认走的是DeepEP的prefill模式，加上混合EP后可以用上DeepEP的low latency模式（decode），bsz越大加速可能更明显一点。 |
RichardWooSJTU
approved these changes
Jul 25, 2025
| num_max_dispatch_tokens_per_rank, | ||
| None, | ||
| num_experts, | ||
| ) |
There was a problem hiding this comment.
这里也需要适配develop 下一个PR这里都兼容一下吧
| elif moe_phase == MoEPhase.PREFILL: | ||
| self.deepep_engine = deep_ep.Buffer( | ||
| # prefill engine | ||
| self.prefill_deepep_engine = deep_ep.Buffer( | ||
| self.group, | ||
| int(5e8), | ||
| 0, | ||
| low_latency_mode=False, |
There was a problem hiding this comment.
mixed模式下，这个low_latency_mode不应该设置成True吗？
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Support mixed ep.
TPOT, OTPS, speed of decode are all improved to some extent at different bsz.