diff --git a/tests/kernels/moe/test_flashinfer.py b/tests/kernels/moe/test_flashinfer.py index 9189c5c641..6329780f9d 100644 --- a/tests/kernels/moe/test_flashinfer.py +++ b/tests/kernels/moe/test_flashinfer.py @@ -623,6 +623,9 @@ def test_unquantized_flashinfer_trtllm_weights_can_be_reprocessed( def _check_flashinfer_ipc_weights(entries, expected, mode, device_index): from vllm.model_executor.model_loader.weight_cache.ipc_loader import IpcModelLoader + from vllm.model_executor.model_loader.weight_cache.protocol import ( + WeightCacheState, + ) from vllm.model_executor.utils import weights_already_processed torch.accelerator.set_device_index(device_index) @@ -632,7 +635,7 @@ def _check_flashinfer_ipc_weights(entries, expected, mode, device_index): ) loader = object.__new__(IpcModelLoader) loader.mode = mode - loader._apply_entries(layer, entries, {}, device_index) + loader._apply_entries(layer, WeightCacheState(entries, {}, {}), device_index) pointers = (layer.w13_weight.data_ptr(), layer.w2_weight.data_ptr()) # Like ipc_cache: a fresh method, only tensor metadata, no _setup_kernel.