Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Kernel] Expand FP8 support to Ampere GPUs using FP8 Marlin #5975

Merged
merged 19 commits into from
Jul 3, 2024
Merged

[Kernel] Expand FP8 support to Ampere GPUs using FP8 Marlin #5975

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
19 commits
Select commit Hold shift + click to select a range
7c51175
Add marlin implementation for fp8 decompress
mgoin Jun 21, 2024
a90212a
It runs! And then crashes
mgoin Jun 21, 2024
9c89164
Merge branch 'upstream-main' into fp8-marlin
mgoin Jun 24, 2024
f61c75f
Latest state, "Invalid __global__ read"
mgoin Jun 24, 2024
0a595a9
Cleanup
mgoin Jun 24, 2024
9c825ad
Fix issues with workspace and add custom packing kernel
mgoin Jun 25, 2024
a14123c
Cleanup cmake debug
mgoin Jun 25, 2024
5b424d9
Add a unit test for packing
mgoin Jun 25, 2024
3c91b98
Everything works!
mgoin Jun 26, 2024
1244ab1
Cleanup
mgoin Jun 26, 2024
0854c81
It works AND it is fast
mgoin Jun 27, 2024
2fc0c47
Add unit test
mgoin Jun 28, 2024
c4b3e0d
Merge branch 'upstream-main' into fp8-marlin
mgoin Jul 1, 2024
abbf3ff
Fix merge
mgoin Jul 1, 2024
ec58535
Fix comment
mgoin Jul 1, 2024
2490f06
Merge branch 'upstream-main' into fp8-marlin
mgoin Jul 2, 2024
ff97364
Merge branch 'upstream-main' into fp8-marlin
mgoin Jul 3, 2024
26ad9bb
Use new platform capability interface
mgoin Jul 3, 2024
aa52a08
Try to work around current_platform issues
mgoin Jul 3, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Next Next commit
Add marlin implementation for fp8 decompress
  • Loading branch information
@mgoin
mgoin committed Jun 21, 2024
commit 7c5117565c6af82f86c73260d63aad4bb1d6000b
1 change: 1 addition & 0 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -178,6 +178,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
"csrc/quantization/marlin/sparse/marlin_24_cuda_kernel.cu"
"csrc/quantization/gptq_marlin/gptq_marlin.cu"
"csrc/quantization/gptq_marlin/gptq_marlin_repack.cu"
"csrc/quantization/fp8/fp8_marlin.cu"
"csrc/custom_all_reduce.cu"
"csrc/quantization/cutlass_w8a8/scaled_mm_entry.cu"
"csrc/quantization/cutlass_w8a8/scaled_mm_c2x.cu"
Expand Down
6 changes: 6 additions & 0 deletions csrc/ops.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,6 +92,12 @@ torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
int64_t size_k, int64_t size_n,
int64_t num_bits);

torch::Tensor fp8_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
torch::Tensor& b_scales, torch::Tensor& g_idx,
torch::Tensor& perm, torch::Tensor& workspace,
int64_t num_bits, int64_t size_m, int64_t size_n,
int64_t size_k, bool is_k_full);

bool cutlass_scaled_mm_supports_fp8(int64_t cuda_device_capability);

void cutlass_scaled_mm(torch::Tensor& out, torch::Tensor const& a,
Expand Down
Loading