[2026春季][T2-1-3] goog00#1390
Open
goog00 wants to merge 1 commit into
Open
Conversation
Signed-off-by: sunteng <steng2009@163.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
概述
为 Qwen3-MoE 适配新增
grouped_gemm算子:一次调用内对多个专家的权重 slab 依次发 GEMM,替代逐专家小 GEMM,用于 MoE experts 前向的批化加速。
改动
grouped_gemm算子:C ABI + Graph API,CPU / NVIDIA / MetaX 三后端 + CPU 参考实现group_sizes直通(默认nullptr,向后兼容,为拷贝-同步受限平台预留优化位)
test/infiniop/grouped_gemm.py(F16/BF16/F32)TensorImpl::zeros在 MetaX 上整型不清零的问题(该后端走 CPU 构造后拷贝,其余后端保持 device 端 memset)
测试
随 InfiniLM
2026-spring-goog00-T2-1-3分支 侧 Qwen3-30B-A3B-Thinking-2507 端到端验证:NVIDIA RTX 4090D(TP=4)+ 沐曦 MetaX C500(TP=2)双平台 prefill 对齐 HF 各 4/4 PASS,
grouped_gemm三后端均验证有效。详见 InfiniLM PR描述。Honor Code.pdf