On the OpenAI API, davinci (or code-davinci-002 if you have access to it still). LLaMA is probably the best model you can run locally. Others you can run locally include EleutherAI's GPT-NeoX-20B, various sizes of Pythia, and GPT-J-6B.
May 2, 2023 at 4:30 PM
On the OpenAI API, davinci (or code-davinci-002 if you have access to it still). LLaMA is probably the best model you can run locally. Others you can run locally include EleutherAI's GPT-NeoX-20B, various sizes of Pythia, and GPT-J-6B.
It will auto-select the closest aspect ratio out of the three Bluesky allows, you then have to manually click the aspect ratio it ended up as in the post interface.
April 14, 2023 at 5:29 PM
It will auto-select the closest aspect ratio out of the three Bluesky allows, you then have to manually click the aspect ratio it ended up as in the post interface.
It makes a .png file that is the same name as your original file except it has _pad at the end. It pads with transparent black by default, if you want opaque white add "--bg 255 255 255 255" to the command line. 🙂
April 14, 2023 at 5:28 PM
It makes a .png file that is the same name as your original file except it has _pad at the end. It pads with transparent black by default, if you want opaque white add "--bg 255 255 255 255" to the command line. 🙂
It's a Python script, you can paste it into an editor or download it by right clicking the "raw" button. You need Pillow installed, which you can do with "pip install pillow" if you don't have it already. Then you run it with "python3 bluesky_pad.py <image_to_pad>".
April 14, 2023 at 5:27 PM
It's a Python script, you can paste it into an editor or download it by right clicking the "raw" button. You need Pillow installed, which you can do with "pip install pillow" if you don't have it already. Then you run it with "python3 bluesky_pad.py <image_to_pad>".