[Fix] FP8 Qwen3-Next quant error by removing fallback fused shards (#23973)

This commit is contained in:
Jimmy Shong
2026-04-29 17:33:47 -04:00
committed by GitHub
parent 850021378a
commit 3d31ac2672
2 changed files with 52 additions and 2 deletions
@@ -60,8 +60,6 @@ def _module_path_match(ignored: str, prefix: str) -> bool:
_FALLBACK_FUSED_SHARDS: Mapping[str, List[str]] = {
"qkv_proj": ["q_proj", "k_proj", "v_proj"],
"gate_up_proj": ["gate_proj", "up_proj"],
"in_proj_ba": ["in_proj_b", "in_proj_a"],
"in_proj_qkvz": ["in_proj_qkv", "in_proj_z"],
}