Skip to content

[ROCm][Bugfix][GPTOSS]: fix input_ids and expert_map args for quark w4a8 gptoss #41165

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
gshtras merged 4 commits into from
Apr 29, 2026
+5 −5
Merged

[ROCm][Bugfix][GPTOSS]: fix input_ids and expert_map args for quark w4a8 gptoss #41165

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
6af0337
fix input_ids and expert_map args for quark w4a8 gptoss
Rohan138 Apr 28, 2026
f9f72aa
fix return and error string
Rohan138 Apr 28, 2026
c9d8623
Merge branch 'main' into fix_gptoss_w4a8_loading
Rohan138 Apr 29, 2026
8c884f0
Merge branch 'main' into fix_gptoss_w4a8_loading
Rohan138 Apr 29, 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
10 changes: 5 additions & 5 deletions vllm/model_executor/layers/quantization/quark/quark_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1577,14 +1577,14 @@ def is_monolithic(self) -> bool:

def apply_monolithic(
self,
layer: torch.nn.Module,
layer: FusedMoE,
x: torch.Tensor,
router_logits: torch.Tensor,
expert_map: torch.Tensor | None = None,
) -> torch.Tensor | tuple[torch.Tensor, torch.Tensor]:
input_ids: torch.Tensor | None = None,
) -> torch.Tensor:
if layer.enable_eplb:
raise NotImplementedError(
"EPLB not supported for `QuarkW4MXFp4MoEMethod_OSS` yet."
f"EPLB not supported for {self.__class__.__name__} yet."
)

from vllm.model_executor.layers.fused_moe.experts.gpt_oss_triton_kernels_moe import ( # noqa: E501
Expand All @@ -1601,7 +1601,7 @@ def apply_monolithic(
topk=layer.top_k,
renormalize=layer.renormalize,
global_num_experts=layer.global_num_experts,
expert_map=expert_map,
expert_map=layer.expert_map,
quant_config=self.moe_quant_config,
apply_router_weight_on_input=layer.apply_router_weight_on_input,
unpadded_N_w1=self.moe.intermediate_size_per_partition_unpadded * 2,
Expand Down
Loading