@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 60d1c6e584
Fork 0

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 16, 19:30 UTC

[Kernel] Fix deprecation function warnings squeezellm quant_cuda_kernel (#6901)

Tyler Michael Smith committed Jul 29, 2024 at 12:59 UTC 60d1c6e584203b07df5020529d678e69baad84a3
1 file changed +2 -2
csrc/quantization/squeezellm/quant_cuda_kernel.cu
+2 -2
@@ -197,13 +197,13 @@ void squeezellm_gemm(torch::Tensor vec, torch::Tensor mat, torch::Tensor mul,
197 const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
198 vllm::squeezellm::NUQ4MatMulKernel<<<blocks, threads, 0, stream>>>(
199 #ifndef USE_ROCM
200 - (half2*)vec.data<at::Half>(),
200 + (half2*)vec.data_ptr<at::Half>(),
201 #else
202 (__half2*)vec.data_ptr<at::Half>(),
203 #endif
204 mat.data_ptr<int>(),
205 #ifndef USE_ROCM
206 - (half2*)mul.data<at::Half>(), (__half*)lookup_table.data<at::Half>(),
206 + (half2*)mul.data<at::Half>(), (__half*)lookup_table.data_ptr<at::Half>(),
207 #else
208 (float2*)mul.data_ptr<float>(),
209 (__half*)lookup_table.data_ptr<at::Half>(),