陈铖豪 chenchenghao
  • Joined on 2026-07-03
chenchenghao pushed to cch/megamoe-rc-v0.1 at chenchenghao/sglang 2026-07-09 08:50:57 +08:00
ebf11be355 fix(moe): gate fp8 megamoe on deepgemm capability
89ba17ad05 feat(moe): add fp8 megamoe fused path and nvfp4 layout guards
eb7f44a8ee fix(moe): pass megamoe fp4 weights as int8 to deepgemm
f482a06a51 test(moe): exercise megamoe a2a handler without dummy shortcut
0d3bf6b5c2 fix(moe): import capture mode from cuda graph runner
Compare 10 commits »
chenchenghao created branch cch/megamoe-rc-v0.1 in chenchenghao/sglang 2026-07-09 08:50:56 +08:00
chenchenghao pushed to cch/deepgemm-glm52-nvfp4-group16 at chenchenghao/DeepGEMM 2026-07-09 08:49:49 +08:00
007c645f87 fix(megamoe): normalize fp4 weight preparation contract
2c7543130b feat(megamoe): expose fp4 weight preparation helper
8ad348fb11 feat(megamoe): add fp4 group16 to group32 requant path
79fcfd6abf feat(megamoe): add nvfp4 group16 capability gate
0ff91f1285 docs: record sm90 megamoe phase6 dev log
Compare 10 commits »
chenchenghao created branch cch/deepgemm-glm52-nvfp4-group16 in chenchenghao/DeepGEMM 2026-07-09 08:49:49 +08:00
chenchenghao created repository chenchenghao/DeepGEMM 2026-07-09 08:49:06 +08:00
chenchenghao created repository chenchenghao/sglang 2026-07-09 08:48:10 +08:00