Re:Show HN: I spent a weekend building a tool that lets you ...
by dylanintech on Tuesday July 25, 2023 @09:52PM
(#36872204)
so think of LoRAs as "extensions" for ML models.
sometimes you need to generate content that's outside of the base model's training set (ex: paintings of your face).
generally speaking there are two options here: to finetune the entire model (very costly in terms of both compute and monetary cost) or to use a LoRA.
usually the process of making a LoRA would include uploading your training images to some directory, executing some python code via either google colabs or running github scripts.
this process can generally be very error-prone and time consuming (even for a programmer like myself) so i built this app to try to make it easier.
so far, 100s of people have used the app and 9 users are monthly subscribers.
sometimes you need to generate content that's outside of the base model's training set (ex: paintings of your face).
generally speaking there are two options here: to finetune the entire model (very costly in terms of both compute and monetary cost) or to use a LoRA.
usually the process of making a LoRA would include uploading your training images to some directory, executing some python code via either google colabs or running github scripts.
this process can generally be very error-prone and time consuming (even for a programmer like myself) so i built this app to try to make it easier.
so far, 100s of people have used the app and 9 users are monthly subscribers.