fix: fatal bugfix in batch decode operator #177
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
yzh119
pushed a commit
that referenced
this pull request
Mar 19, 2025
Compiling FlashInfer on CUDA 12.2 triggers errors such as those shown below. This PR aims to fix this issue. ``` #10 1380.2 /usr/flexflow-serve/deps/flashinfer/include/flashinfer/attention/../vec_dtypes.cuh(55): error: more than one instance of function "flashinfer::__hmul" matches the argument list: #10 1380.2 function "__hmul(__nv_bfloat16, __nv_bfloat16)" (declared at line 3518 of /usr/local/cuda/include/cuda_bf16.hpp) #10 1380.2 function "flashinfer::__hmul(__nv_bfloat16, __nv_bfloat16)" (declared at line 44) #10 1380.2 argument types are: (const __nv_bfloat16, const __nv_bfloat16) #10 1380.2 val.x = __hmul(a.x, b.x); #10 1380.2 ^ #10 1380.2 #10 1380.2 /usr/flexflow-serve/deps/flashinfer/include/flashinfer/attention/../vec_dtypes.cuh(56): error: more than one instance of function "flashinfer::__hmul" matches the argument list: #10 1380.2 function "__hmul(__nv_bfloat16, __nv_bfloat16)" (declared at line 3518 of /usr/local/cuda/include/cuda_bf16.hpp) #10 1380.2 function "flashinfer::__hmul(__nv_bfloat16, __nv_bfloat16)" (declared at line 44) #10 1380.2 argument types are: (const __nv_bfloat16, const __nv_bfloat16) #10 1380.2 val.y = __hmul(a.y, b.y); #10 1380.2 ^ #10 1380.2 #10 1380.2 /usr/flexflow-serve/deps/flashinfer/include/flashinfer/attention/../vec_dtypes.cuh(1158): error: more than one instance of function "flashinfer::make_bfloat162" matches the argument list: #10 1380.2 function "make_bfloat162(__nv_bfloat16, __nv_bfloat16)" (declared at line 1189 of /usr/local/cuda/include/cuda_bf16.hpp) #10 1380.2 function "flashinfer::make_bfloat162(__nv_bfloat16, __nv_bfloat16)" (declared at line 37) #10 1380.2 argument types are: (nv_bfloat16, nv_bfloat16) #10 1380.2 data = make_bfloat162(val, val); #10 1380.2 ^ #10 1380.2 #10 1380.2 /usr/flexflow-serve/deps/flashinfer/include/flashinfer/attention/../vec_dtypes.cuh(1203): error: more than one instance of function "flashinfer::make_bfloat162" matches the argument list: #10 1380.2 function "make_bfloat162(__nv_bfloat16, __nv_bfloat16)" (declared at line 1189 of /usr/local/cuda/include/cuda_bf16.hpp) #10 1380.2 function "flashinfer::make_bfloat162(__nv_bfloat16, __nv_bfloat16)" (declared at line 37) #10 1380.2 argument types are: (nv_bfloat16, nv_bfloat16) #10 1380.2 *(nv_bfloat162*)(&data.x) = make_bfloat162(val, val); #10 1380.2 ^ #10 1380.2 #10 1380.2 /usr/flexflow-serve/deps/flashinfer/include/flashinfer/attention/../vec_dtypes.cuh(1204): error: more than one instance of function "flashinfer::make_bfloat162" matches the argument list: #10 1380.2 function "make_bfloat162(__nv_bfloat16, __nv_bfloat16)" (declared at line 1189 of /usr/local/cuda/include/cuda_bf16.hpp) #10 1380.2 function "flashinfer::make_bfloat162(__nv_bfloat16, __nv_bfloat16)" (declared at line 37) #10 1380.2 argument types are: (nv_bfloat16, nv_bfloat16) #10 1380.2 *(nv_bfloat162*)(&data.y) = make_bfloat162(val, val); #10 1380.2 ^ #10 1380.2 #10 1384.6 /usr/flexflow-serve/deps/flashinfer/include/flashinfer/attention/../mma.cuh(524): warning #177-D: variable "s_u32" was declared but never referenced #10 1384.6 uint32_t* s_u32 = (uint32_t*)(s); #10 1384.6 ^ #10 1384.6 detected during: #10 1384.6 instantiation of "void flashinfer::<unnamed>::compute_sfm_v<KTraits>(flashinfer::smem_t<KTraits::SWIZZLE_MODE_KV> *, uint32_t *, KTraits::DTypeQKAccum (*)[KTraits::NUM_MMA_KV][8], float (*)[KTraits::NUM_MMA_D_VO][8], float (*)[2]) [with KTraits=flashinfer::KernelTraits<flashinfer::MaskMode::kNone, 128U, 2U, 8U, 4U, 4U, 4U, 1U, flashinfer::PosEncodingMode::kNone, half, half, half, float, int32_t, flashinfer::DefaultAttention<false, false, false, false>>]" at line 2127 of /usr/flexflow-serve/deps/flashinfer/include/flashinfer/attention/prefill.cuh ```
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The
BatchDecodeWithPagedKVCacheWrapperdidn't run into the kernel.