minor updates

Signed-off-by: Sage Moore <[email protected]>
vllm-project · Dec 9, 2024 · 6d4b8d0 · 6d4b8d0
1 parent 6915fa2
commit 6d4b8d0
Show file tree

Hide file tree

Showing 2 changed files with 1 addition and 3 deletions.
diff --git a/tests/kernels/test_fused_quant_activation.py b/tests/kernels/test_fused_quant_activation.py
@@ -16,9 +16,7 @@
 
 def ref_impl(silu_and_mul: SiluAndMul, x: torch.Tensor,
              scale: torch.Tensor) -> torch.Tensor:
-    # Norm
     silu_and_mul_out = silu_and_mul.forward_native(x)
-    # Quant
     out, scales = ops.scaled_fp8_quant(silu_and_mul_out, scale)
     return out
 

diff --git a/vllm/compilation/pass_manager.py b/vllm/compilation/pass_manager.py
@@ -45,7 +45,7 @@ def configure(self, pass_config: CompilationConfig.PassConfig):
         if pass_config.enable_reshape:
             self.passes += [RedundantReshapesPass(pass_config)]
 
-        if True:
+        if pass_config.enable_fusion:
             self.passes += [FusionPass.instance(pass_config)]
             self.passes += [ActivationQuantFusionPass.instance(pass_config)]