Closed sirredbeard closed 5 days ago
The file also contains a typo in:
To just try try the base model without fine-tuning you can run this command after activating conda.
This appears to be done in first_time_setup.sh
but is not executing, or needs to be documented that the user should run.
For local finetuning, AI Toolkit will download the model raw files and execute first_time_setup.sh
when generating the project. E.g.,
(After downloading, it executes first_time_setup.sh
)
Do you see any similar output from the log panel, or any error message?
BTW, good suggestion on the doc and I'll update the document in case user needs to run that manually.
@swatDong I did not see that on first run, but I set up another fine tuning project, it downloaded the models, and attempted to re-run the first_time_setup.sh
. Re-running the script failed then because I already had a conda env with the same name, but it was simple enough to pick it up from there. I am not sure what happened on the first run, may have been user error.
Thank you for updating the doc.
When performing local fine tuning, this Preview README.md appears:
I cannot find the text of this file in the repository, unlike the others for remote fine tuning, so I cannot open a PR to suggest fixes.
However, it is containing missing steps.
base
conda environment, not the one specified in the project setup page.conda activate [conda-env-name]
but you need to runconda env create -f setup/conda-environment.yml
first, otherwise conda doesn't see the environment specified in the project setup page.