Summary:
I have an issue when running the PaliGemma block, the block returns an OOM Killed error
Expected behavior:
The core block, should receive the inputs which are the hf_token, the image that needs analysis, and the prompt of what needs to be analysed
Actual behavior:
Takes approximately a run time of 5 minutes, and fails to provide output image, which is caused by an OOM Killed Error
Steps to reproduce:
- Drop Processing block on canvas
- Connect the password block to the hf_token, and provide a hf_token as an input parameter
- Connect a file upload block, and pass the desired image for analysis into the file upload block
- Connect a text block, and pass in the text prompt on what the analysis is supposed to be
- Connect a view image block, For receiving output of the processing paligemma block.
Zetaforge version:
0.3.2
Operating System
MacOs
Other operating system:
No response
Logs:
[2024/10/29 17:40:48][hjxzw-pali-gemma-20wmu5ij3mfl] Cannot authenticate through git-credential as no helper is defined on your machine.
[2024/10/29 17:40:48][hjxzw-pali-gemma-20wmu5ij3mfl] You might have to re-authenticate when pushing to the Hugging Face Hub.
[2024/10/29 17:40:48][hjxzw-pali-gemma-20wmu5ij3mfl] Run the following command in your terminal in case you want to set the 'store' credential helper as default.
[2024/10/29 17:40:48][hjxzw-pali-gemma-20wmu5ij3mfl] git config --global credential.helper store
[2024/10/29 17:40:48][hjxzw-pali-gemma-20wmu5ij3mfl] Read https://git-scm.com/book/en/v2/Git-Tools-Credential-Storage for more details.
[2024/10/29 17:40:48][hjxzw-pali-gemma-20wmu5ij3mfl] Token has not been saved to git credential helper.
[2024/10/29 17:40:49][hjxzw-pali-gemma-20wmu5ij3mfl] /usr/local/lib/python3.10/dist-packages/transformers/models/paligemma/configuration_paligemma.py:137: FutureWarning: The `vocab_size` attribute is deprecated and will be removed in v4.44, Please use `text_config.vocab_size` instead.
[2024/10/29 17:40:49][hjxzw-pali-gemma-20wmu5ij3mfl] warnings.warn(
[2024/10/29 17:48:11][hjxzw-pali-gemma-20wmu5ij3mfl]
Downloading shards: 0%| | 0/3 [00:00<?, ?it/s]
Downloading shards: 33%|███▎ | 1/3 [03:22<06:45, 202.68s/it]
Downloading shards: 67%|██████▋ | 2/3 [06:24<03:10, 190.49s/it]
Downloading shards: 100%|██████████| 3/3 [07:22<00:00, 129.77s/it]
Downloading shards: 100%|██████████| 3/3 [07:22<00:00, 147.38s/it]
[2024/10/29 17:48:12][hjxzw-pali-gemma-20wmu5ij3mfl] `config.hidden_act` is ignored, you should use `config.hidden_activation` instead.
[2024/10/29 17:48:12][hjxzw-pali-gemma-20wmu5ij3mfl] Gemma's activation function will be set to `gelu_pytorch_tanh`. Please, use
[2024/10/29 17:48:12][hjxzw-pali-gemma-20wmu5ij3mfl] `config.hidden_activation` if you want to override this behaviour.
[2024/10/29 17:48:12][hjxzw-pali-gemma-20wmu5ij3mfl] See https://github.com/huggingface/transformers/pull/29402 for more details.
[2024/10/29 17:48:12][hjxzw-pali-gemma-20wmu5ij3mfl]
Loading checkpoint shards: 0%| | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards: 33%|███▎ | 1/3 [00:00<00:00, 4.94it/s]
Loading checkpoint shards: 100%|██████████| 3/3 [00:00<00:00, 8.63it/s]
Loading checkpoint shards: 100%|██████████| 3/3 [00:00<00:00, 8.02it/s]
[2024/10/29 17:48:19] error during pipeline execution; err=rpc error: code = Unauthenticated desc = Unauthorized
Recording:
No response
Anything else?:
No response
Summary:
I have an issue when running the PaliGemma block, the block returns an OOM Killed error
Expected behavior:
The core block, should receive the inputs which are the hf_token, the image that needs analysis, and the prompt of what needs to be analysed
Actual behavior:
Takes approximately a run time of 5 minutes, and fails to provide output image, which is caused by an OOM Killed Error
Steps to reproduce:
Zetaforge version:
0.3.2
Operating System
MacOs
Other operating system:
No response
Logs:
Recording:
No response
Anything else?:
No response