[ROCm][Bugfix][GPTOSS]: fix input_ids and expert_map args for quark w4a8 gptoss (#41165)
Signed-off-by: Rohan138 <rohanpotdar138@gmail.com>
This commit is contained in:
@@ -1573,14 +1573,14 @@ class QuarkOCP_MX_MoEMethod_OSS(QuarkOCP_MX_MoEMethod):
|
|||||||
|
|
||||||
def apply_monolithic(
|
def apply_monolithic(
|
||||||
self,
|
self,
|
||||||
layer: torch.nn.Module,
|
layer: FusedMoE,
|
||||||
x: torch.Tensor,
|
x: torch.Tensor,
|
||||||
router_logits: torch.Tensor,
|
router_logits: torch.Tensor,
|
||||||
expert_map: torch.Tensor | None = None,
|
input_ids: torch.Tensor | None = None,
|
||||||
) -> torch.Tensor | tuple[torch.Tensor, torch.Tensor]:
|
) -> torch.Tensor:
|
||||||
if layer.enable_eplb:
|
if layer.enable_eplb:
|
||||||
raise NotImplementedError(
|
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
|
from vllm.model_executor.layers.fused_moe.experts.gpt_oss_triton_kernels_moe import ( # noqa: E501
|
||||||
@@ -1597,7 +1597,7 @@ class QuarkOCP_MX_MoEMethod_OSS(QuarkOCP_MX_MoEMethod):
|
|||||||
topk=layer.top_k,
|
topk=layer.top_k,
|
||||||
renormalize=layer.renormalize,
|
renormalize=layer.renormalize,
|
||||||
global_num_experts=layer.global_num_experts,
|
global_num_experts=layer.global_num_experts,
|
||||||
expert_map=expert_map,
|
expert_map=layer.expert_map,
|
||||||
quant_config=self.moe_quant_config,
|
quant_config=self.moe_quant_config,
|
||||||
apply_router_weight_on_input=layer.apply_router_weight_on_input,
|
apply_router_weight_on_input=layer.apply_router_weight_on_input,
|
||||||
unpadded_N_w1=self.moe.intermediate_size_per_partition_unpadded * 2,
|
unpadded_N_w1=self.moe.intermediate_size_per_partition_unpadded * 2,
|
||||||
|
|||||||
Reference in New Issue
Block a user