diff --git a/tests/unittest/_torch/attention/test_attention_op_sync.py b/tests/unittest/_torch/attention/test_attention_op_sync.py index f85667b5a85e..3aeec7fa00a9 100644 --- a/tests/unittest/_torch/attention/test_attention_op_sync.py +++ b/tests/unittest/_torch/attention/test_attention_op_sync.py @@ -63,7 +63,6 @@ } _THOP_KWARG_SOURCE_ALIASES: dict[str, tuple[str, tuple[str, ...]]] = { - "beam_width": ("metadata", ("effective_beam_width",)), "context_lengths": ("metadata", ("prompt_lens_cuda_runtime",)), "head_size": ("attn", ("head_dim",)), "host_context_lengths": ("metadata", ("prompt_lens_cpu_runtime",)), @@ -74,6 +73,14 @@ "metadata", ("max_total_draft_tokens",), ), + "skip_softmax_threshold_scale_factor_decode": ( + "skip_softmax_kernel_params", + ("threshold_scale_factor_decode",), + ), + "skip_softmax_threshold_scale_factor_prefill": ( + "skip_softmax_kernel_params", + ("threshold_scale_factor_prefill",), + ), "workspace_": ("metadata", ("effective_workspace",)), }