[Kernel] Support deep_gemm for linear methods (#19085)

Signed-off-by: artetaout <lulala341@gmail.com>
This commit is contained in:
artetaout
2025-06-11 15:14:45 +08:00
committed by GitHub
parent 5039ec2336
commit b8e809a057
3 changed files with 124 additions and 1 deletions
@@ -402,6 +402,7 @@ class Fp8LinearMethod(LinearMethodBase):
if self.block_quant:
assert self.quant_config.weight_block_size is not None
return torch.ops.vllm.apply_w8a8_block_fp8_linear(
input=x,
weight=layer.weight,