ImportError: cannot import name 'tiled_flexible_dual_grid_to_mesh' from 'o_voxel.convert'

[grapefruitlips]

Hi, where can I find the latest version of o_voxel to build it from source?

There is no tiled_flexible_dual_grid_to_mesh function in https://github.com/microsoft/TRELLIS.2/blob/main/o-voxel/o_voxel/convert/flexible_dual_grid.py

[visualbruno]

Did you install the last wheels about o_voxel and cumesh ?
I updated them in the last commit

[grapefruitlips]

i tried to build everything from source because I have python3.13, I couldn't find weels for cp313

[visualbruno]

You can find the source I used

• o_voxel : https://github.com/visualbruno/Trellis.2
• FlexGEMM : https://github.com/visualbruno/FlexGEMM
• CuMesh : https://github.com/visualbruno/CuMesh

[master-merlin]

Also Pytorch 2.9.1 is missing which is very commonly used on 50xx with ComfyUI.

Can be found here - but as it seems they aren't compatible for soem reason with the recently implemented tiled approach as after the update from yesterday I am not able to create models anymore: https://github.com/PozzettiAndrea/ovoxel-wheels

[visualbruno]

Yes, about o_voxel "tiled" approach, it's a custom code I made

[visualbruno]

About o_voxel "tiled" approach, you can use the same version of o_voxel you have until now, because I did not change the compiled code, but only the file that is in the folder "o_voxel\convert\flexible_dual_grid.py"

For me, it is saved in the virtual environment of python here: venv\Lib\site-packages\o_voxel\convert\

[master-merlin]

Wouldn't it be easier to create a custom lib for the "tiled" approach and keep o-voxel close to the original for compatibility and maintenance?

[visualbruno]

My version of o_voxel is compatible with the original one, because I just added a function that does not exist in the original code, so you should not have any problem to use it with another node or project.

[grapefruitlips]

You can find the source I used

• o_voxel : https://github.com/visualbruno/Trellis.2
• FlexGEMM : https://github.com/visualbruno/FlexGEMM
• CuMesh : https://github.com/visualbruno/CuMesh

Thanks! Now everything works fine with cuda_13.0.r13.0 and torch 2.9.1+cu130

For those who will clone https://github.com/visualbruno/TRELLIS.2.git to build o-voxel from source, don't foget to copy eigen (https://github.com/PX4/eigen.git) to TRELLIS.2/o-voxel/third_party/eigen

[master-merlin]

About o_voxel "tiled" approach, you can use the same version of o_voxel you have until now, because I did not change the compiled code, but only the file that is in the folder "o_voxel\convert\flexible_dual_grid.py"

For me, it is saved in the virtual environment of python here: venv\Lib\site-packages\o_voxel\convert\

Confirmed working for Python 3.12 PyTorch 2.9.1 Cuda 12.8

[jason-mightynice]

About o_voxel "tiled" approach, you can use the same version of o_voxel you have until now, because I did not change the compiled code, but only the file that is in the folder "o_voxel\convert\flexible_dual_grid.py"

For me, it is saved in the virtual environment of python here: venv\Lib\site-packages\o_voxel\convert\

I did this change, but now get a different error.

Windows py3.13 torch 2.91 cu12.8

UPDATE: it works again. I just have to disable/uninstall xformers ..

Loading Image Cond model ... Loading Sparse Structure model ... Sampling sparse structure: 100%|███████████████████████████████████████████████████████| 20/20 [00:03<00:00, 5.22it/s] Loading Shape Slat Flow 512 model ... Loading Shape Slat Flow 1024 model ... Sampling shape SLat (LR): 100%|████████████████████████████████████████████████████████| 20/20 [00:04<00:00, 4.04it/s] Loading Shape Slat decoder model ... Sampling shape SLat (HR): 100%|████████████████████████████████████████████████████████| 20/20 [01:08<00:00, 3.44s/it] Loading Texture Slat Flow 1024 model ... Sampling texture SLat: 100%|███████████████████████████████████████████████████████████| 20/20 [00:53<00:00, 2.70s/it] !!! Exception during processing !!! max(): Expected reduction dim 0 to have non-zero size. Traceback (most recent call last): File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\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 "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\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 "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\execution.py", line 303, in _async_map_node_over_list await process_inputs(input_dict, i) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\execution.py", line 291, in process_inputs result = f(**inputs) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\nodes.py", line 841, 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 "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\utils\_contextlib.py", line 120, in decorate_context return func(*args, **kwargs) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\pipelines\trellis2_image_to_3d.py", line 1116, in run out_mesh = self.decode_latent(shape_slat, tex_slat, res, use_tiled=use_tiled) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\utils\_contextlib.py", line 120, in decorate_context return func(*args, **kwargs) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\pipelines\trellis2_image_to_3d.py", line 796, in decode_latent meshes, subs = self.decode_shape_slat(shape_slat, resolution, use_tiled=use_tiled) ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\pipelines\trellis2_image_to_3d.py", line 684, in decode_shape_slat ret = self.models['shape_slat_decoder'](slat, return_subs=True, useTiled=use_tiled) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\models\sc_vaes\fdg_vae.py", line 85, in forward decoded = super().forward(x, **kwargs) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\models\sc_vaes\sparse_unet_vae.py", line 605, in forward h, sub = block(h) ~~~~~^^^ File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\models\sc_vaes\sparse_unet_vae.py", line 335, in forward return self._forward(x, subdiv) ~~~~~~~~~~~~~^^^^^^^^^^^ File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\models\sc_vaes\sparse_unet_vae.py", line 324, in _forward h = self.conv2(h) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1775, in _wrapped_call_impl return self._call_impl(*args, **kwargs) ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1786, in _call_impl return forward_call(*args, **kwargs) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\modules\sparse\conv\conv.py", line 19, in forward return _backends[config.CONV].sparse_conv3d_forward(self, x) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^ File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\modules\sparse\conv\conv_flex_gemm.py", line 49, in sparse_conv3d_forward torch.Size([*x.shape, *x.spatial_shape]), ^^^^^^^^^^^^^^^ File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\modules\sparse\basic.py", line 494, in spatial_shape spatial_shape = self.__cal_spatial_shape(self.coords) File "D:\ComfyUI_windows_latest\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Trellis2_VB\trellis2\modules\sparse\basic.py", line 474, in __cal_spatial_shape return torch.Size((coords[:, 1:].max(0)[0] + 1).tolist()) ~~~~~~~~~~~~~~~~~^^^ IndexError: max(): Expected reduction dim 0 to have non-zero size.

[JuyeongYi]

if you want to use latest pytorch, you need to build every torch-related packages and usually py313 is not a good idea.

[jason-mightynice]

if you want to use latest pytorch, you need to build every torch-related packages and usually py313 is not a good idea.

yeah, I keep a 3.12 environment handy, but haven't needed it for a while. Also just went from a 4090 to 5090 which hasn't been the smoothest!