mxfp8 online quant move to new frontend (#40152)

Signed-off-by: Vasiliy Kuznetsov <vasiliy@meta.com>
This commit is contained in:
Vasiliy Kuznetsov
2026-04-20 06:26:12 -07:00
committed by GitHub
parent a023edfa5b
commit 38fa87caca
7 changed files with 108 additions and 126 deletions
@@ -515,14 +515,6 @@ class Fp8OnlineLinearMethod(Fp8LinearMethod):
initialize_online_processing(layer)
# TODO: remove this check once the following RFC is resolved.
# https://github.com/vllm-project/vllm/issues/33314
# Subclasses (e.g. Mxfp8OnlineLinearMethod) only need the weight
# registration above and manage their own kernel, so skip fp8_linear
# kernel creation for them.
if type(self) is not Fp8OnlineLinearMethod:
return
self.fp8_linear = init_fp8_linear_kernel(
activation_quant_key=self.activation_quant_key,
weight_quant_key=self.weight_quant_key,