DispatchStub: missing kernel for cuda

[HerzogVonWiesel]

Hey there, sadly I get the following error when I try to run the workflow:

got prompt
Loading Image Cond model ...
Loading weights: 100%|████████████████████████████| 415/415 [00:00<00:00, 3803.31it/s, Materializing param=norm.weight]
Loading Sparse Structure model ...
[ATTENTION] Using backend: xformers
Sampling sparse structure: 100%|███████████████████████████████████████████████████████| 12/12 [00:02<00:00,  4.78it/s]
Loading Shape Slat Flow 512 model ...
Loading Shape Slat Flow 1024 model ...
Sampling shape SLat (LR):   0%|                                                                 | 0/12 [00:00<?, ?it/s]
!!! Exception during processing !!! false INTERNAL ASSERT FAILED at "C:\\actions-runner\\_work\\pytorch\\pytorch\\pytorch\\aten\\src\\ATen\\native\\DispatchStub.cpp":276, please report a bug to PyTorch. DispatchStub: missing kernel for cuda
Traceback (most recent call last):
  File "C:\_AI\ComfyUI\execution.py", line 518, 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 "C:\_AI\ComfyUI\execution.py", line 329, 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 "C:\_AI\ComfyUI\execution.py", line 303, in _async_map_node_over_list
    await process_inputs(input_dict, i)
  File "C:\_AI\ComfyUI\execution.py", line 291, in process_inputs
    result = f(**inputs)
             ^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\nodes.py", line 313, in process
    mesh = pipeline.run(image=image_in, seed=seed, pipeline_type=pipeline_type, sparse_structure_sampler_params = sparse_structure_sampler_params, shape_slat_sampler_params = shape_slat_sampler_params, tex_slat_sampler_params = tex_slat_sampler_params, max_num_tokens = max_num_tokens, sparse_structure_resolution = sparse_structure_resolution, max_views = max_views, generate_texture_slat = generate_texture_slat, use_tiled=use_tiled_decoder, pbar=pbar)[0]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\venv\Lib\site-packages\torch\utils\_contextlib.py", line 120, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\pipelines\trellis2_image_to_3d.py", line 992, in run
    shape_slat, res = self.sample_shape_slat_cascade(
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\pipelines\trellis2_image_to_3d.py", line 576, in sample_shape_slat_cascade
    slat = self.shape_slat_sampler.sample(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\venv\Lib\site-packages\torch\utils\_contextlib.py", line 120, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\pipelines\samplers\flow_euler.py", line 208, in sample
    return super().sample(model, noise, cond, steps, rescale_t, verbose, neg_cond=neg_cond, guidance_strength=guidance_strength, guidance_interval=guidance_interval, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\venv\Lib\site-packages\torch\utils\_contextlib.py", line 120, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\pipelines\samplers\flow_euler.py", line 121, in sample
    out = self.sample_once(model, sample, t, t_prev, cond, **kwargs)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\venv\Lib\site-packages\torch\utils\_contextlib.py", line 120, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\pipelines\samplers\flow_euler.py", line 79, in sample_once
    pred_x_0, pred_eps, pred_v = self._get_model_prediction(model, x_t, t, cond, **kwargs)
                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\pipelines\samplers\flow_euler.py", line 49, in _get_model_prediction
    pred_v = self._inference_model(model, x_t, t, cond, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\pipelines\samplers\guidance_interval_mixin.py", line 11, in _inference_model
    return super()._inference_model(model, x_t, t, cond, guidance_strength=guidance_strength, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\pipelines\samplers\classifier_free_guidance_mixin.py", line 23, in _inference_model
    std_pos = x_0_pos.std(dim=list(range(1, x_0_pos.ndim)), keepdim=True)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\modules\sparse\basic.py", line 296, in std
    mean = self.mean(dim=dim, keepdim=True)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\modules\sparse\basic.py", line 287, in mean
    return self.reduce(op='mean', dim=dim, keepdim=keepdim)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\_AI\ComfyUI\custom_nodes\ComfyUI-Trellis2\trellis2\modules\sparse\basic.py", line 283, in reduce
    red = torch.segment_reduce(red, reduce=op, lengths=self.seqlen)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: false INTERNAL ASSERT FAILED at "C:\\actions-runner\\_work\\pytorch\\pytorch\\pytorch\\aten\\src\\ATen\\native\\DispatchStub.cpp":276, please report a bug to PyTorch. DispatchStub: missing kernel for cuda

I'm not sure what's going wrong exactly, since it seems to have downloaded the models itself, as well as me having put the facebook model in the correct place. Normal image generation (ZIT, WAN 2.2) work without issue.

Thank you in advance!

Windows 11, AMD 7950X, Python 3.12
Total VRAM 24564 MB, total RAM 64734 MB
pytorch version: 2.8.0+cu129
xformers version: 0.0.32.post2
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 4090 : cudaMallocAsync
Using async weight offloading with 2 streams

[easy664049283-creator]

Some outdated 3D methods are missing in cu129. Try reverting to cu128

[HerzogVonWiesel]

My entire environment relies on cu129 (xformers, torch, etc.), so I'd rather not have to change it. Especially since Trellis seems to run just fine on cu129 according to https://forums.developer.nvidia.com/t/trellis-2-3d-mode-gen-from-microsoft/355030/19 ?