@@ -2736,7 +2736,8 @@ def test_w4_1gpu(self, moe_backend, cuda_graph, overlap_scheduler, mocker):
2736
2736
with llm :
2737
2737
model_name = "GPT-OSS/MXFP4"
2738
2738
mocker .patch .object (GSM8K , "MAX_OUTPUT_LEN" , 8192 )
2739
- mocker .patch .dict (GSM8K .EVALUATE_KWARGS , {"scores_filter" : "exact_match,flexible-extract" })
2739
+ mocker .patch .dict (GSM8K .EVALUATE_KWARGS ,
2740
+ {"scores_filter" : "exact_match,flexible-extract" })
2740
2741
task = GSM8K (model_name )
2741
2742
task .evaluate (llm ,
2742
2743
extra_evaluator_kwargs = self .extra_evaluator_kwargs )
@@ -2778,7 +2779,8 @@ def test_w4_4gpus(self, moe_backend, tp_size, pp_size, ep_size,
2778
2779
model_name = "GPT-OSS/MXFP4"
2779
2780
task = GSM8K (model_name )
2780
2781
mocker .patch .object (GSM8K , "MAX_OUTPUT_LEN" , 8192 )
2781
- mocker .patch .dict (GSM8K .EVALUATE_KWARGS , {"scores_filter" : "exact_match,flexible-extract" })
2782
+ mocker .patch .dict (GSM8K .EVALUATE_KWARGS ,
2783
+ {"scores_filter" : "exact_match,flexible-extract" })
2782
2784
task .evaluate (llm ,
2783
2785
extra_evaluator_kwargs = self .extra_evaluator_kwargs )
2784
2786
@@ -2810,7 +2812,8 @@ def test_w4a16(self, tp_size, pp_size, ep_size, attention_dp, cuda_graph,
2810
2812
model_name = "GPT-OSS/BF16"
2811
2813
task = GSM8K (model_name )
2812
2814
mocker .patch .object (GSM8K , "MAX_OUTPUT_LEN" , 8192 )
2813
- mocker .patch .dict (GSM8K .EVALUATE_KWARGS , {"scores_filter" : "exact_match,flexible-extract" })
2815
+ mocker .patch .dict (GSM8K .EVALUATE_KWARGS ,
2816
+ {"scores_filter" : "exact_match,flexible-extract" })
2814
2817
task .evaluate (llm ,
2815
2818
extra_evaluator_kwargs = self .extra_evaluator_kwargs )
2816
2819
0 commit comments