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, 18:15 UTC
[AMD][CI/Build] Disambiguation of the function call for ROCm 6.2 headers compatibility (#7477)
Co-authored-by: Charlie Fu <Charlie.Fu@amd.com>
Gregory Shtrasberg committed
Aug 21, 2024 at 19:47 UTC
9984605412de1171a72d955cfcb954725edd4d6f
1 file changed
+1
-1
csrc/attention/attention_utils.cuh
+1
-1
@@ -34,7 +34,7 @@ inline __device__ float qk_dot_(const Vec (&q)[N], const Vec (&k)[N]) {
34
A_vec qk_vec = mul<A_vec, Vec, Vec>(q[0], k[0]);
35
#pragma unroll
36
for (int ii = 1; ii < N; ++ii) {
37
- qk_vec = fma(q[ii], k[ii], qk_vec);
37
+ qk_vec = vllm::fma(q[ii], k[ii], qk_vec);
38
}
39
40
// Finalize the reduction across lanes.