[SYCL] supprt Flash Attention for fp32/fp16/Q4/Q5/Q8 (#20190)

* support flash-attention for fp32/fp16/Q4/Q5/Q8

* rm warining

* update for JIT
This commit is contained in:
Neo Zhang
2026-03-08 12:00:07 +08:00
committed by GitHub
parent c5a778891b
commit 213c4a0b81
65 changed files with 20091 additions and 8593 deletions
+16 -1
View File
@@ -29,6 +29,21 @@ using to_t_nc_sycl_t = void (*)(const void * x, T * y, int64_t ne00, int64_t ne0
int64_t s01, int64_t s02, int64_t s03, dpct::queue_ptr queue);
typedef to_t_nc_sycl_t<sycl::half> to_fp16_nc_sycl_t;
to_fp16_nc_sycl_t get_to_fp16_nc_sycl(ggml_type type);
to_fp16_nc_sycl_t ggml_get_to_fp16_nc_sycl(ggml_type type);
template<typename dst_t, typename src_t>
inline dst_t ggml_sycl_cast(src_t x) {
if constexpr (std::is_same_v<dst_t, src_t>) {
return x;
} else if constexpr (std::is_same_v<dst_t, sycl::ext::oneapi::bfloat16>) {
return sycl::ext::oneapi::bfloat16(float(x));
} else if constexpr (std::is_same_v<src_t, sycl::ext::oneapi::bfloat16>) {
return static_cast<float>(x);
} else if constexpr(std::is_same_v<dst_t, int32_t>) {
return int32_t(x);
} else {
return float(x);
}
}
#endif // GGML_SYCL_CONVERT_HPP