A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 19:45 UTC
[Bugfix][W8A8] fixed cutlass block fp8 binding (#14796)
DefTruth committed
Mar 14, 2025 at 18:32 UTC
40253bab443ad0cdd22ff33bd8f777d2f289cfc4
1 file changed
+1
-1
csrc/torch_bindings.cpp
+1
-1
@@ -370,7 +370,7 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
370
"cutlass_scaled_mm_supports_block_fp8(int cuda_device_capability) -> "
371
"bool");
372
ops.impl("cutlass_scaled_mm_supports_block_fp8",
373
- &cutlass_scaled_mm_supports_fp8);
373
+ &cutlass_scaled_mm_supports_block_fp8);
374
375
// Check if cutlass sparse scaled_mm is supported for CUDA devices of the
376
// given capability