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, 08:30 UTC
[Bugfix] Add half type support in reshape_and_cache_cpu_impl on x86 cpu platform (#18430)
Signed-off-by: Yuqi Zhang <yuqizhang@google.com> Co-authored-by: Yuqi Zhang <yuqizhang@google.com>
Yuqi Zhang committed
May 23, 2025 at 01:41 UTC
d0bc2f810b7a34247154b078c2429bf62519e9ca
1 file changed
+1
csrc/cpu/cpu_types_x86.hpp
+1
@@ -19,6 +19,7 @@ namespace vec_op {
19
#define VLLM_DISPATCH_CASE_FLOATING_TYPES_FP8(...) \
20
AT_DISPATCH_CASE(at::ScalarType::Float, __VA_ARGS__) \
21
AT_DISPATCH_CASE(at::ScalarType::BFloat16, __VA_ARGS__) \
22
+ AT_DISPATCH_CASE(at::ScalarType::Half, __VA_ARGS__) \
23
AT_DISPATCH_CASE(at::ScalarType::Float8_e5m2, __VA_ARGS__)
24
25
#define VLLM_DISPATCH_FLOATING_TYPES(TYPE, NAME, ...) \