Update compressed_tensors_w8a8_fp8.py

HabanaAI · Sep 25, 2024 · 5726801 · 5726801
1 parent 3e8762e
commit 5726801
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/...el_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_fp8.py b/...el_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_fp8.py
@@ -23,8 +23,8 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
     def __init__(self, strategy: str, is_static_input_scheme: bool):
         self.strategy = strategy
         self.is_static_input_scheme = is_static_input_scheme
-        self.cutlass_fp8_supported = cutlass_fp8_supported(
-        ) if torch.cuda.is_available() else False
+        self.cutlass_fp8_supported = not current_platform.is_hpu() and \
+                                     cutlass_fp8_supported()
 
     @classmethod
     def get_min_capability(cls) -> int: