syy

i use deforum-art 3d zoom out + img2img + controlnet to make video encountered a problem: black area of background becomes bigger and bigger https://storage.googleapis.com/mlserver-environment/1684773335.mp4

question about c++ deploy stable difussion

2

hi, i am using c++ api to deploy stable difussion，I encountered two problems 1. i have set constant using code ` SetConstant(model_handler_, 'output_0', &ait_output); ` but got error when run...

bug for train magvit2

16

## ❓ Questions and Help code：https://github.com/lucidrains/magvit2-pytorch train script： ``` from magvit2_pytorch import ( VideoTokenizer, VideoTokenizerTrainer ) tokenizer = VideoTokenizer( image_size = 128, codebook_size=1024, init_dim = 64, max_dim = 512, layers...

question about img2img

2

请问T2I-Adapter支持img2img + sdxl + cannyAdapter吗？ 后续有机会做成diffusers的pipeline吗？

recon images is black

9

i train magvit2 on imagenet the recon images is black, but the loss looks normal  

Distributed training is stuck

4

### Description I use two 4090 host for data parallel distributed training by jax.distributed, like this: jax.distributed.initialize(coordinator_address="[ip]:[port]", num_processes=2, process_id=[index]) the train is stuck when doing all_reduce ops  How...

how to add a custom dataset

1

how to add a custom dataset like ucf101