Skip to content

Revert "[AMD] Fix RotaryEmbedding crash on AMD/ROCm (regression from #17934)" #18922

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
HaiShaw merged 1 commit into from
Feb 17, 2026
Merged

Revert "[AMD] Fix RotaryEmbedding crash on AMD/ROCm (regression from #17934)" #18922

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
13 changes: 0 additions & 13 deletions python/sglang/srt/layers/rotary_embedding.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -389,19 +389,6 @@ def forward_cuda(
)
return query, key

def forward_hip(self, *args, **kwargs):
"""HIP/ROCm implementation.

The JIT kernels (sglang.jit_kernel.pos_enc) used in forward_cuda's
fallback path depend on tvm_ffi which invokes nvidia-smi to detect
CUDA compute capability. This fails on AMD GPUs, so we use the
pure-PyTorch native implementation instead.

Uses *args/**kwargs because subclasses (MRotaryEmbedding, etc.)
have different forward_native() signatures.
"""
return self.forward_native(*args, **kwargs)

def extra_repr(self) -> str:
s = f"head_size={self.head_size}, rotary_dim={self.rotary_dim}"
s += f", max_position_embeddings={self.max_position_embeddings}"
Expand Down
Loading