Skip to content

Revert "perf(ROCm): add is_rdna() detection and optimize CE loss for RDNA GPUs" #4139

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

Merged
danielhanchen merged 2 commits into from
Mar 3, 2026
Merged

Revert "perf(ROCm): add is_rdna() detection and optimize CE loss for RDNA GPUs" #4139

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ed297ae
Revert "perf(ROCm): optimize chunked CE loss num_warps for RDNA GPUs …
GoldenGrapeGentleman Mar 2, 2026
380fb35
Merge branch 'main' into revert-4123-perf/rdna-kernel-tuning
GoldenGrapeGentleman Mar 3, 2026
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
3 changes: 1 addition & 2 deletions unsloth/kernels/cross_entropy_loss.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,6 @@
triton_cast,
torch_gpu_device,
is_cdna,
is_rdna,
)
from transformers.models.llama.modeling_llama import logger
from unsloth_zoo.utils import Version
Expand Down Expand Up @@ -365,7 +364,7 @@ def forward(
SOFTCAP = logit_softcapping,
DO_LOGIT_SCALING = DO_LOGIT_SCALING,
LOGIT_SCALE = logit_scaling,
num_warps = 16 if is_cdna() or is_rdna() else 32,
num_warps = 32 if not is_cdna() else 16,
)
# logsumexp(chunked_logsumexp) - x
# Do the -x separately
Expand Down