As a student in an introductory university course on digital history, I tried to recreate the code and analytical operations of "Interrogating a National Narrative with GPT-2" in the Google Colab environment using T4 CPU a few weeks ago. Unfortunately, I was unsuccessful in executing it.
I suspect that the aitextgen library might have become deprecated (the latest version was released on August 9, 2022). I tried downgrading the versions of the libraries aitextgen depends on, but could not find a version of pytorch-lightening that runs in Python 3.12 (which is the Python version that Colab currently employs) and met the requirements of aitextgen:
- If I specified to use a pytorch-lightning version that is older than 1.6.0., I encountered the following error: "AttributeError: module 'pkgutil' has no attribute 'ImpImporter'".
- If I specified to use a pytorch-lightning version that is at least 1.6.0 or newer, I encountered another error: "MisconfigurationException: The provided lr scheduler
LambdaLR doesn't follow PyTorch's LRScheduler API. You should override the LightningModule.lr_scheduler_step hook with your own logic if you are using a custom LR scheduler."
Since my knowledge of Python is very limited, I was not able to follow the suggested solution in the error code, and it is also possible that there is an easy workaround that I have missed. You can access my code, which includes annotations, via this link: https://colab.research.google.com/drive/1yUlk0tCN9ZbQhGG0QpTP3ifo56c_l02d?usp=sharing
Additionally, I am including my documentation produced for the course.
Documentation-GPT_Language_Models_as_Lenses.pdf
Thank you very much for providing and maintaining so many valuable tutorials that are of great use in my learning journey on digital history!
As a student in an introductory university course on digital history, I tried to recreate the code and analytical operations of "Interrogating a National Narrative with GPT-2" in the Google Colab environment using T4 CPU a few weeks ago. Unfortunately, I was unsuccessful in executing it.
I suspect that the aitextgen library might have become deprecated (the latest version was released on August 9, 2022). I tried downgrading the versions of the libraries aitextgen depends on, but could not find a version of pytorch-lightening that runs in Python 3.12 (which is the Python version that Colab currently employs) and met the requirements of aitextgen:
LambdaLRdoesn't follow PyTorch's LRScheduler API. You should override theLightningModule.lr_scheduler_stephook with your own logic if you are using a custom LR scheduler."Since my knowledge of Python is very limited, I was not able to follow the suggested solution in the error code, and it is also possible that there is an easy workaround that I have missed. You can access my code, which includes annotations, via this link: https://colab.research.google.com/drive/1yUlk0tCN9ZbQhGG0QpTP3ifo56c_l02d?usp=sharing
Additionally, I am including my documentation produced for the course.
Documentation-GPT_Language_Models_as_Lenses.pdf
Thank you very much for providing and maintaining so many valuable tutorials that are of great use in my learning journey on digital history!