Did you fine tune Anima's llm_adapter?
#8
by Disya - opened
Hi! I wanted to ask about the text-conditioning part of Anima-Light-Lavender.
During training, did you also fine-tune Anima's llm_adapter, or was it kept frozen and only the DiT was trained?
I'm specifically asking about the conditioning adapter after Qwen, not the Qwen model itself.
Thanks!
it was trained jointly with the transformer during the first half of training, and frozen for the second half.
Oh, that explains why the trigger words from the lora - anima base, don't work with this model