LLM inference in C/C++
C++ ggml
HIP: bump HIP_VERSION requried for fp8 to avoid missing __hip_fp8_e4m3 support in 6.2 (#29231)
uvosd028c697b529
diff --git a/ggml/src/ggml-cuda/vendors/hip.h b/ggml/src/ggml-cuda/vendors/hip.hindex 0a2f2829e5..da7e3da0b9 100644--- a/ggml/src/ggml-cuda/vendors/hip.h+++ b/ggml/src/ggml-cuda/vendors/hip.h@@ -248,11 +248,11 @@typedef __hip_bfloat16 nv_bfloat16;typedef __hip_bfloat162 nv_bfloat162;-#if HIP_VERSION >= 60200000+#if HIP_VERSION >= 60300000#include <hip/hip_fp8.h>typedef __hip_fp8_e4m3 __nv_fp8_e4m3;#define FP8_AVAILABLE-#endif // HIP_VERSION >= 60200000+#endif // HIP_VERSION >= 60300000typedef int8_t int8x4_t __attribute__((ext_vector_type(4)));typedef uint8_t uint8x4_t __attribute__((ext_vector_type(4)));