[Quantization] Add ModelOpt NVFP4 W4A16 (4-bit weights, fp16/bf16 activations) support (#41769)

Signed-off-by: Juhi Mittal <juhim@nvidia.com>
This commit is contained in:
Juhi Mittal
2026-05-09 21:15:50 +00:00
committed by GitHub
parent 2ee8c2a56e
commit 7a2b596982
3 changed files with 224 additions and 6 deletions
+1
View File
@@ -60,6 +60,7 @@ WEIGHT_LOADER_V2_SUPPORTED = [
"ModelOptFp8PbWoLinearMethod",
"QuarkLinearMethod",
"ModelOptNvFp4LinearMethod",
"ModelOptNvFp4W4A16LinearMethod",
"HummingLinearMethod",
]