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] Fix CUTLASS 3.x custom broadcast load epilogue #5516

Merged
merged 3 commits into from
Jun 14, 2024
Merged

[Kernel] Fix CUTLASS 3.x custom broadcast load epilogue #5516

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
87160e5
Fix broadcast load epilogue
tlrmchlsmth Jun 13, 2024
bf129c8
turn on cutlass again
tlrmchlsmth Jun 13, 2024
b25b3f3
format
tlrmchlsmth Jun 13, 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
2 changes: 1 addition & 1 deletion csrc/quantization/cutlass_w8a8/broadcast_load_epilogue_c3x.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -153,7 +153,7 @@ struct Sm90RowOrScalarBroadcast {

CUTLASS_DEVICE void
begin(uint64_t* full_mbarrier_ptr, int load_iteration, bool issue_tma_load) {
if (params.ptr_row == nullptr) {
if (!params.row_broadcast) {
return;
}

Expand Down
4 changes: 1 addition & 3 deletions vllm/model_executor/layers/quantization/fp8.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -257,9 +257,7 @@ def apply(self,
# If dynamic, layer.input_scale is None and x_scale computed from x.
# If static, layer.input_scale is scalar and x_scale is input_scale.

# Temporarily disable CUTLASS kernels due to an illegal memory access
#if bias is None and self.cutlass_fp8_supported:
if False:
if bias is None and self.cutlass_fp8_supported:
qinput, x_scale = ops.scaled_fp8_quant(x, layer.input_scale)

# Fused GEMM_DQ
Expand Down
Loading