Merged lora vs. lora Loader
#52
by Fronyax - opened
If I merge this lora directly into the Krea 2 model itself, can I expect it to behave the same way as when applying it with the lora loader?
If I merge this lora directly into the Krea 2 model itself, can I expect it to behave the same way as when applying it with the lora loader?
Merging is basically what's done in inference; that is, if you merge this LoRa into the base model, you can train a LoRa on that model. The big issue is that in inference, you would need to have the merged base model being used with your LoRa. Another way would be to use this LoRa as a base to initialize the weights of your LoRa and train something new on top of that.