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 8, 03:45 UTC
[BugFix] [Build] fix string literals comparison in indexer_k_quant_and_cache calling site (#34653)
Signed-off-by: Hongxia Yang <hongxiay.yang@amd.com> Co-authored-by: Hongxia Yang <hongxiay.yang@amd.com>
Hongxia Yang committed
Feb 17, 2026 at 22:19 UTC
4a00a511bbf707fcb484d655b3b5eec0ed0ca308
1 file changed
+2
-1
csrc/cache_kernels.cu
+2
-1
@@ -1305,7 +1305,8 @@ void indexer_k_quant_and_cache(
1305
const at::cuda::OptionalCUDAGuard device_guard(device_of(k));
1306
const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
1307
1308
- DISPATCH_BY_KV_CACHE_DTYPE(k.dtype(), "fp8_e4m3",
1308
+ static const std::string kv_cache_dtype = "fp8_e4m3";
1309
+ DISPATCH_BY_KV_CACHE_DTYPE(k.dtype(), kv_cache_dtype,
1310
CALL_INDEXER_K_QUANT_AND_CACHE);
1311
}
1312