Bagheera

so i can actually hit this one on CUDA, **but** it only happens during training without autocast 🤔 the weights are in bf16 precision, not fp32. maybe this is what...

i'm constantly updating my models on the hub and this behaviour means that the latest model is always loaded for inference. disabling the network for the user when they have...

the base model was trained using it, so i figured aligning with the base model's training and inference has better results. from my own tests, i can now reduce the...

see the base SDXL pipeline: ```py # get unconditional embeddings for classifier free guidance zero_out_negative_prompt = negative_prompt is None and self.config.force_zeros_for_empty_prompt if do_classifier_free_guidance and negative_prompt_embeds is None and zero_out_negative_prompt: negative_prompt_embeds...

can i also open the pull request for all of the other training examples, to add general dropout capabilities to them?

like the ticket for updating the fp16 error? #6231

a good demonstration of the current generation of models' two primary forms of residual noise would probably be a good idea though i can't think of how to integrate that....

wow, i hadn't expected all of those, or really looked into them until now. i'm not sure why this simple check made those tests fail. i am wondering if there's...

i was working on this support but after seeing the results of the model i'm not sure it's ready to be added yet:  there's a lot of residual noise...

@JincanDeng how are you doing caption dropout? zeroes or `""` prompt encoded by both TEs?