Skip to content

GitLab

Commit dd823f7f authored by zhuwenwen's avatar zhuwenwen
Browse files

add bf16

parent 9fed1f5d
Hide whitespace changes
Inline Side-by-side
Showing with 8 additions and 8 deletions
csrc/cache_kernels.cu
View file @ dd823f7f
...@@ -251,8 +251,8 @@ void reshape_and_cache( ...@@ -251,8 +251,8 @@ void reshape_and_cache(
CALL_RESHAPE_AND_CACHE(float, float, false); CALL_RESHAPE_AND_CACHE(float, float, false);
} else if (key.dtype() == at::ScalarType::Half) { } else if (key.dtype() == at::ScalarType::Half) {
CALL_RESHAPE_AND_CACHE(uint16_t, uint16_t, false); CALL_RESHAPE_AND_CACHE(uint16_t, uint16_t, false);
// } else if (key.dtype() == at::ScalarType::BFloat16) { } else if (key.dtype() == at::ScalarType::BFloat16) {
// CALL_RESHAPE_AND_CACHE(__nv_bfloat16, __nv_bfloat16, false); CALL_RESHAPE_AND_CACHE(__nv_bfloat16, __nv_bfloat16, false);
} }
// } else if (kv_cache_dtype == "fp8_e5m2") { // } else if (kv_cache_dtype == "fp8_e5m2") {
// if (key.dtype() == at::ScalarType::Float) { // if (key.dtype() == at::ScalarType::Float) {
...@@ -308,8 +308,8 @@ void convert_fp8_e5m2( ...@@ -308,8 +308,8 @@ void convert_fp8_e5m2(
CALL_CONVERT_FP8_E5M2(uint8_t, float); CALL_CONVERT_FP8_E5M2(uint8_t, float);
} else if (src_cache.dtype() == at::ScalarType::Half) { } else if (src_cache.dtype() == at::ScalarType::Half) {
CALL_CONVERT_FP8_E5M2(uint8_t, uint16_t); CALL_CONVERT_FP8_E5M2(uint8_t, uint16_t);
// } else if (src_cache.dtype() == at::ScalarType::BFloat16) { } else if (src_cache.dtype() == at::ScalarType::BFloat16) {
// CALL_CONVERT_FP8_E5M2(uint8_t, __nv_bfloat16); CALL_CONVERT_FP8_E5M2(uint8_t, __nv_bfloat16);
} else if (dst_cache.dtype() == at::ScalarType::Float) { } else if (dst_cache.dtype() == at::ScalarType::Float) {
CALL_CONVERT_FP8_E5M2(float, uint8_t); CALL_CONVERT_FP8_E5M2(float, uint8_t);
} else if (dst_cache.dtype() == at::ScalarType::Half) { } else if (dst_cache.dtype() == at::ScalarType::Half) {
......
csrc/dispatch_utils.h
View file @ dd823f7f
...@@ -8,8 +8,8 @@ ...@@ -8,8 +8,8 @@
#define VLLM_DISPATCH_CASE_FLOATING_TYPES(...) \ #define VLLM_DISPATCH_CASE_FLOATING_TYPES(...) \
AT_DISPATCH_CASE(at::ScalarType::Float, __VA_ARGS__) \ AT_DISPATCH_CASE(at::ScalarType::Float, __VA_ARGS__) \
AT_DISPATCH_CASE(at::ScalarType::Half, __VA_ARGS__) AT_DISPATCH_CASE(at::ScalarType::Half, __VA_ARGS__) \
// AT_DISPATCH_CASE(at::ScalarType::BFloat16, __VA_ARGS__) AT_DISPATCH_CASE(at::ScalarType::BFloat16, __VA_ARGS__)
#define VLLM_DISPATCH_FLOATING_TYPES(TYPE, NAME, ...) \ #define VLLM_DISPATCH_FLOATING_TYPES(TYPE, NAME, ...) \
AT_DISPATCH_SWITCH( \ AT_DISPATCH_SWITCH( \
...@@ -17,8 +17,8 @@ ...@@ -17,8 +17,8 @@
#define VLLM_DISPATCH_CASE_FLOATING_AND_BYTE_TYPES(...) \ #define VLLM_DISPATCH_CASE_FLOATING_AND_BYTE_TYPES(...) \
AT_DISPATCH_CASE(at::ScalarType::Float, __VA_ARGS__) \ AT_DISPATCH_CASE(at::ScalarType::Float, __VA_ARGS__) \
AT_DISPATCH_CASE(at::ScalarType::Half, __VA_ARGS__) AT_DISPATCH_CASE(at::ScalarType::Half, __VA_ARGS__) \
// AT_DISPATCH_CASE(at::ScalarType::BFloat16, __VA_ARGS__) \ AT_DISPATCH_CASE(at::ScalarType::BFloat16, __VA_ARGS__)
// AT_DISPATCH_CASE(at::ScalarType::Byte, __VA_ARGS__) // AT_DISPATCH_CASE(at::ScalarType::Byte, __VA_ARGS__)
#define VLLM_DISPATCH_FLOATING_AND_BYTE_TYPES(TYPE, NAME, ...) \ #define VLLM_DISPATCH_FLOATING_AND_BYTE_TYPES(TYPE, NAME, ...) \
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment