RuntimeError: CUDA error: invalid configuration argument CUDA kernel errors might be asynchronously reported at some other API call,

[xiatiantuo]

RuntimeError: CUDA error: invalid configuration argument
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
这是什么错误 ？

[smthemex]

呃~多给点信息，这个我也不清楚是插件问题，还是插件冲突，或者cuda环境的问题

[xiatiantuo]

呃~多给点信息，这个我也不清楚是插件问题，还是插件冲突，或者cuda环境的问题

RuntimeError: CUDA error: invalid configuration argument
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

以前的老版本怎么没了？老版本挺好用的

File "/root/ComfyUI/execution.py", line 525, in execute
output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

File "/root/ComfyUI/execution.py", line 334, in get_output_data
return_values = await _async_map_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

File "/root/ComfyUI/execution.py", line 308, in _async_map_node_over_list
await process_inputs(input_dict, i)

File "/root/ComfyUI/execution.py", line 296, in process_inputs
result = f(**inputs)

File "/root/ComfyUI/comfy_api/internal/init.py", line 149, in wrapped_func
return method(locked_class, **inputs)

File "/root/ComfyUI/comfy_api/latest/_io.py", line 1764, in EXECUTE_NORMALIZED
to_return = cls.execute(*args, **kwargs)

File "/root/ComfyUI/custom_nodes/ComfyUI_DiffuEraser/diffueraser_node.py", line 221, in execute
image_list=model.forward( copy.deepcopy(conditioning["video_image"]), copy.deepcopy(conditioning["init_mask"]), copy.deepcopy(conditioning["prioris"]),folder_paths.get_output_directory(),positive,load_videobypath=False,

File "/root/ComfyUI/custom_nodes/ComfyUI_DiffuEraser/libs/diffueraser.py", line 461, in forward
images = self.pipeline(

File "/root/ComfyUI/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)

File "/root/ComfyUI/custom_nodes/ComfyUI_DiffuEraser/libs/pipeline_diffueraser.py", line 1277, in call
noise_pred = self.unet(

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
return forward_call(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/diffusers/hooks/hooks.py", line 189, in new_forward
output = function_reference.forward(*args, **kwargs)

File "/root/ComfyUI/custom_nodes/ComfyUI_DiffuEraser/libs/unet_motion_model.py", line 845, in forward
sample, res_samples = downsample_block(

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
return forward_call(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/diffusers/hooks/hooks.py", line 189, in new_forward
output = function_reference.forward(*args, **kwargs)

File "/root/ComfyUI/custom_nodes/ComfyUI_DiffuEraser/libs/unet_3d_blocks.py", line 1252, in forward
hidden_states = motion_module(

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
return forward_call(*args, **kwargs)

File "/root/ComfyUI/custom_nodes/ComfyUI_DiffuEraser/libs/transformer_temporal.py", line 175, in forward
hidden_states = block(

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
return forward_call(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/diffusers/models/attention.py", line 995, in forward
attn_output = self.attn1(

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
return forward_call(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/diffusers/models/attention_processor.py", line 605, in forward
return self.processor(

File "/root/ComfyUI/lib/python3.10/site-packages/diffusers/models/attention_processor.py", line 2558, in call
hidden_states = xformers.ops.memory_efficient_attention(

File "/root/ComfyUI/lib/python3.10/site-packages/xformers/ops/fmha/init.py", line 306, in memory_efficient_attention
return _memory_efficient_attention(

File "/root/ComfyUI/lib/python3.10/site-packages/xformers/ops/fmha/init.py", line 467, in _memory_efficient_attention
return _memory_efficient_attention_forward(

File "/root/ComfyUI/lib/python3.10/site-packages/xformers/ops/fmha/init.py", line 490, in memory_efficient_attention_forward
out, * = op.apply(inp, needs_gradient=False)

File "/root/ComfyUI/lib/python3.10/site-packages/xformers/ops/fmha/flash.py", line 677, in apply
out, softmax_lse, rng_state = cls.OPERATOR(

File "/root/ComfyUI/lib/python3.10/site-packages/torch/_ops.py", line 1158, in call
return self._op(*args, **(kwargs or {}))

File "/root/ComfyUI/lib/python3.10/site-packages/torch/_library/custom_ops.py", line 335, in backend_impl
result = self._backend_fns[device_type](*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/torch/_compile.py", line 51, in inner
return disable_fn(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/torch/_dynamo/eval_frame.py", line 838, in _fn
return fn(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/torch/_library/custom_ops.py", line 367, in wrapped_fn
return fn(*args, **kwargs)

File "/root/ComfyUI/lib/python3.10/site-packages/xformers/ops/fmha/flash.py", line 151, in _flash_fwd
out, softmax_lse, p, rng_state = _C_flashattention.fwd(

[smthemex]

torch/_dynamo 这个问题，好像是最近才出来的，之前的没有，你的torch版本是多少，以及cuda版本， 因为你这个报错 我也看着眼熟

[xiatiantuo]

Please 'pip install apex'Total VRAM 24083 MB, total RAM 515791 MB pytorch version: 2.7.0+cu128 xformers version: 0.0.30 Set vram state to: NORMAL_VRAM Device: cuda:0 NVIDIA GeForce RTX 4090 D : cudaMallocAsync ***@***.***> 在 2026年3月27日 周五 19:31 写道： smthemex left a comment (smthemex/ComfyUI_DiffuEraser#38) torch/_dynamo 这个问题，好像是最近才出来的，之前的没有，你的torch版本是多少，以及cuda版本， 因为你这个报错 我也看着眼熟 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.

[smthemex]

我跑一下看看， 按理不应该 但是comfyUI最近更新了内存优化技术

[xiatiantuo]

我将代码改一下可以用，但出来画质变低清了 ***@***.***> 在 2026年3月27日 周五 21:16 写道： smthemex left a comment (smthemex/ComfyUI_DiffuEraser#38) 我跑一下看看， 按理不应该 但是comfyUI最近更新了内存优化技术 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.