OOM

[littleloulou]

你好，我使用786X1024的图片作为底图，使用“workflow_FUN_I2V_GGUF_Q4_0.png”这个工作流，生成1024分辨率的视频会爆显存；768分辨率下没问题。
显卡：RTX 4060TI 16GB.
OS: Ubuntu22.04 Docker 中运行
vae encode tiling 已经打开

是不是因为我的底图太大了？

[wailovet]

有没有报错信息 我想知道在哪一段代码爆的显存

[wailovet]

不过我自己跑的话,很极限,1024占用会在99%, 你可以尝试减少一下视频帧长度

[littleloulou]

有没有报错信息 我想知道在哪一段代码爆的显存

这是报错信息：

ComfyUI Error Report

Error Details

• Node Type: CogVideoXFunSampler
• Exception Type: torch.OutOfMemoryError
• Exception Message: Allocation on device

Stack Trace

  File "/root/ComfyUI/execution.py", line 323, in execute
    output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)

  File "/root/ComfyUI/execution.py", line 198, in get_output_data
    return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)

  File "/root/ComfyUI/execution.py", line 169, in _map_node_over_list
    process_inputs(input_dict, i)

  File "/root/ComfyUI/execution.py", line 158, in process_inputs
    results.append(getattr(obj, func)(**inputs))

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoXWrapper/nodes.py", line 1024, in process
    latents = pipe(

  File "/root/miniconda3/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoX-MZ/cogvideox_fun/pipeline_cogvideox_inpaint.py", line 725, in __call__
    _, masked_video_latents = self.prepare_mask_latents(

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoX-MZ/cogvideox_fun/pipeline_cogvideox_inpaint.py", line 346, in prepare_mask_latents
    mask_pixel_values_bs = self.vae.encode(mask_pixel_values_bs)[0]

  File "/root/miniconda3/lib/python3.10/site-packages/diffusers/utils/accelerate_utils.py", line 46, in wrapper
    return method(self, *args, **kwargs)

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoX-MZ/mz_enable_vae_encode_tiling.py", line 29, in encode
    h = self._encode(x)

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoX-MZ/mz_enable_vae_encode_tiling.py", line 116, in _encode
    return self.tiled_encode(x)

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoX-MZ/mz_enable_vae_encode_tiling.py", line 85, in tiled_encode
    tile = self.encoder(tile)

  File "/root/miniconda3/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1553, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)

  File "/root/miniconda3/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
    return forward_call(*args, **kwargs)

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoX-MZ/cogvideox_fun/autoencoder_magvit.py", line 733, in forward
    hidden_states = down_block(hidden_states, temb, None)

  File "/root/miniconda3/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1553, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)

  File "/root/miniconda3/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
    return forward_call(*args, **kwargs)

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoX-MZ/cogvideox_fun/autoencoder_magvit.py", line 409, in forward
    hidden_states = resnet(hidden_states, temb, zq)

  File "/root/miniconda3/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1553, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)

  File "/root/miniconda3/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
    return forward_call(*args, **kwargs)

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoX-MZ/cogvideox_fun/autoencoder_magvit.py", line 291, in forward
    hidden_states = self.conv1(hidden_states)

  File "/root/miniconda3/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1553, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)

  File "/root/miniconda3/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1562, in _call_impl
    return forward_call(*args, **kwargs)

  File "/root/ComfyUI/custom_nodes/ComfyUI-CogVideoX-MZ/cogvideox_fun/autoencoder_magvit.py", line 142, in forward
    inputs = F.pad(inputs, padding_2d, mode="constant", value=0)

  File "/root/miniconda3/lib/python3.10/site-packages/torch/nn/functional.py", line 4552, in pad
    return torch._C._nn.pad(input, pad, mode, value)

System Information

• ComfyUI Version: v0.2.2-18-g54fca4a
• Arguments: main.py --port 6006 --listen 0.0.0.0
• OS: posix
• Python Version: 3.10.8 (main, Nov 24 2022, 14:13:03) [GCC 11.2.0]
• Embedded Python: false
• PyTorch Version: 2.4.1+cu121

Devices

• Name: cuda:0 NVIDIA GeForce RTX 4060 Ti : cudaMallocAsync
  • Type: cuda
  • VRAM Total: 17175150592
  • VRAM Free: 13030017848
  • Torch VRAM Total: 2986344448
  • Torch VRAM Free: 127290168

[wailovet]

vae encode时的报错,是不是本来跑的时候显存就已经占用了一些了, 我和你一样16G显存,开启vae encode tiling以后将将能用

[littleloulou]

vae encode时的报错,是不是本来跑的时候显存就已经占用了一些了, 我和你一样16G显存,开启vae encode tiling以后将将能用

请问下你是在Linux下使用的吗？我是在win11下使用wsl2 启动的linux子系统，并且通过docker跑的服务，会不会有影响。

[wailovet]

是的,我是直接本机的linux,跑的时候显存占用顶到99%

[littleloulou]

是的,我是直接本机的linux,跑的时候显存占用顶到99%

我装了一个ubuntu22.04，把环境都搞了一遍，960 分辨率都跑不出来，可能是我的原始图片分辨率太高了