hunyuan3D 21 TexGen node hanging

[werksone]

whenever using the Hunyuan 2.1 workflows the hunyuan3D 21 TexGen node hits max GPU/VRAM and then goes no where. does this need more than 24gb of vram?? the model processes fine but it never gets through to the texture generation and just hangs/locks up

using a 4090 and have no other dramas with other models?

[DenisKochetov]

@werksone please try to restart comfy and run only texture mode on a mesh without mmgp optimization

[werksone]

2025-12-02T22:38:13.004135 - !!! Exception during processing !!! Allocation on device 
2025-12-02T22:38:13.058365 - Traceback (most recent call last):
  File "D:\dev\comfy3dStandalone\execution.py", line 510, 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, hidden_inputs=hidden_inputs)
                                                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\execution.py", line 324, 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, hidden_inputs=hidden_inputs)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\execution.py", line 298, in _async_map_node_over_list
    await process_inputs(input_dict, i)
  File "D:\dev\comfy3dStandalone\execution.py", line 286, in process_inputs
    result = f(**inputs)
             ^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\custom_nodes\ComfyUI-3D-Pack\nodes.py", line 5528, in generate
    result_path = texgen_pipe(
                  ^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\custom_nodes\ComfyUI-3D-Pack\Gen_3D_Modules\Hunyuan3D_2_1\hy3dpaint\textureGenPipeline.py", line 152, in __call__
    multiviews_pbr = self.models["multiview_model"](
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\custom_nodes\ComfyUI-3D-Pack\Gen_3D_Modules\Hunyuan3D_2_1\hy3dpaint\utils\multiview_utils.py", line 68, in __call__
    pils = self.forward_one(
           ^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\custom_nodes\ComfyUI-3D-Pack\Gen_3D_Modules\Hunyuan3D_2_1\hy3dpaint\utils\multiview_utils.py", line 113, in forward_one
    mvd_image = self.pipeline(
                ^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\utils\_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\sam.tack\.cache\huggingface\modules\diffusers_modules\local\pipeline.py", line 329, in __call__
    images = self.denoise(
             ^^^^^^^^^^^^^
  File "C:\Users\sam.tack\.cache\huggingface\modules\diffusers_modules\local\pipeline.py", line 717, in denoise
    image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False, generator=generator)[0]
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\diffusers\utils\accelerate_utils.py", line 46, in wrapper
    return method(self, *args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\diffusers\models\autoencoders\autoencoder_kl.py", line 326, in decode
    decoded = self._decode(z).sample
              ^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\diffusers\models\autoencoders\autoencoder_kl.py", line 297, in _decode
    dec = self.decoder(z)
          ^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\module.py", line 1751, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\module.py", line 1762, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\diffusers\models\autoencoders\vae.py", line 337, in forward
    sample = up_block(sample, latent_embeds)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\module.py", line 1751, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\module.py", line 1762, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\diffusers\models\unets\unet_2d_blocks.py", line 2750, in forward
    hidden_states = upsampler(hidden_states)
                    ^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\module.py", line 1751, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\module.py", line 1762, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\diffusers\models\upsampling.py", line 180, in forward
    hidden_states = self.conv(hidden_states)
                    ^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\module.py", line 1751, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\module.py", line 1762, in _call_impl
    return forward_call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\conv.py", line 554, in forward
    return self._conv_forward(input, self.weight, self.bias)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\dev\comfy3dStandalone\venv\Lib\site-packages\torch\nn\modules\conv.py", line 549, in _conv_forward
    return F.conv2d(
           ^^^^^^^^^
torch.OutOfMemoryError: Allocation on device 

2025-12-02T22:38:13.060365 - Got an OOM, unloading all loaded models.
2025-12-02T22:38:13.063365 - Prompt executed in 339.43 seconds

OOM error, feels like 4090 should be able to handle it though without running out of vram, maybe I underestimate it?

[DenisKochetov]

@werksone Hey! I tested the 3D-Pack on a fresh 4090 setup and got the same error.
Reinstalling the libraries fixed it for me:

pip install --force-reinstall "diffusers==0.30.0" "transformers==4.46.0" "accelerate==1.1.1"

With MMGP optimization enabled, everything fits within 24 GB VRAM.
Thanks for testing — looks like we also need to fix the accelerate version in the requirements.