Fooocus Inpaint/Input Image showing Attribute Error

[trialsnabhi]

Read Troubleshoot

[x] I admit that I have read the Troubleshoot before making this issue.

Describe the problem
When I use Inpainting on Fooocus, Iam getting this error offlate. Hence unable to use inpaint on Fooocus at all !

Full Console Log

Traceback (most recent call last):
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/gradio/routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/gradio/blocks.py", line 1429, in process_api
inputs = self.preprocess_data(fn_index, inputs, state)
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/gradio/blocks.py", line 1239, in preprocess_data
processed_input.append(block.preprocess(inputs[i]))
File "/home/studio-lab-user/Fooocus-Sagemaker-Studio-Lab/Fooocus/modules/gradio_hijack.py", line 278, in preprocess
im = processing_utils.decode_base64_to_image(x)
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/gradio/processing_utils.py", line 59, in decode_base64_to_image
img = Image.open(BytesIO(base64.b64decode(image_encoded)))
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/Image.py", line 3288, in open
"""
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/Image.py", line 3274, in _open_core
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/JpegImagePlugin.py", line 799, in jpeg_factory
im = JpegImageFile(fp, filename)
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/ImageFile.py", line 116, in init
self._open()
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/JpegImagePlugin.py", line 380, in _open
handler(self, i)
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/JpegImagePlugin.py", line 211, in SOF
self.mode = "RGB"
AttributeError: can't set attribute 'mode'
Traceback (most recent call last):
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/gradio/routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/gradio/blocks.py", line 1429, in process_api
inputs = self.preprocess_data(fn_index, inputs, state)
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/gradio/blocks.py", line 1239, in preprocess_data
processed_input.append(block.preprocess(inputs[i]))
File "/home/studio-lab-user/Fooocus-Sagemaker-Studio-Lab/Fooocus/modules/gradio_hijack.py", line 278, in preprocess
im = processing_utils.decode_base64_to_image(x)
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/gradio/processing_utils.py", line 59, in decode_base64_to_image
img = Image.open(BytesIO(base64.b64decode(image_encoded)))
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/Image.py", line 3288, in open
"""
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/Image.py", line 3274, in _open_core
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/JpegImagePlugin.py", line 799, in jpeg_factory
im = JpegImageFile(fp, filename)
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/ImageFile.py", line 116, in init
self._open()
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/JpegImagePlugin.py", line 380, in _open
handler(self, i)
File "/home/studio-lab-user/.conda/envs/fooocus/lib/python3.10/site-packages/PIL/JpegImagePlugin.py", line 211, in SOF
self.mode = "RGB"
AttributeError: can't set attribute 'mode'

[trialsnabhi]

It happens when there is any sort of Input Image .. Whether an inpaint/outpaint or ImagePrompt too

[mashb1t]

@trialsnabhi i've checked inpainting in all modes with various jpg images and have not been able to reproduce the issue.
The issue occurs in def SOF(self, marker) of PIL JpegImagePlugin.py, where the layers are mapped and the Start of frame marker can not set the mode accordingly as in your case the object doesn't have such a property.

This is not an issue of Fooocus but most likely a parsing problem of your specific image through PIL.
Can you please attach your source jpg image to debug the layers and / or type of jpg? Thanks!

[trialsnabhi]

oh this happens with any image tat I use in the image prompt/inpatinting source.. either generated image or any jpg/png . Thats why I posted.
To give you more info, I use fooocus on amazonsagemaker labs and kaggle. Both are getting into a similar issue.

[mashb1t]

Interesting... I don't have access to those services, maybe somebody else can support here.

[dhruvchandel]

Facing Similar Issue. I was also using Sagemaker notebook to run Fooocus.

[trialsnabhi]

This may be caused by an indirect early import of PIL in launch.py. The import preceds an installation of Pillow==9.2.0 and remains cached.

If Pillow has been downgraded in this way from a version >=10.1.0, the mixture of modules is obviously incompatible: python-pillow/Pillow#7307

can you help me with a solution ?

[sehajbenipal]

Similar Issue. I was also using Sagemaker and kaggle notebook to run Fooocus.

[mashb1t]

I think the matter has been more or less resolved,. The solution is to either use a notebook without loading libraries before loading Fooocus or restarting the instance afterwards.
Closing for now, but still feel free to provide your feedback.

[mashb1t]

@rsl8 sure, feel free