phyllispeng123

Generated image in blur while training Autoencodertiny Decoder

3

Hi, this is Phyllis. Very appreciate your marvelous work on SD-VAE compression and acceleration !!!! I am currently working on Autoencodertiny decoder training from scratch using LDM training structure, but...

Bad Flux lora training result

7

Hi@kohya-ss , thank you for your detailed and excellent work on FLUX finetuning and lora training!! I got bad results when I ran the sample lora_training script with network_dim=32, input...

Replace DDIM with FlowDPMSolverMultistepScheduler

2

Thanks for you great work, I wonder if the estimate equation also works for FlowDPMSolverMultistepScheduler?

``` from performer_pytorch import FastAttention import torch.nn.functional as F attn_fn = FastAttention( dim_heads = 64, causal=False, no_projection=False) attn_fn_2 = FastAttention( dim_heads = 64, causal=False, no_projection=True) query = torch.randn(2,10,4096,64).to('cuda') key =...

Inference error

1

Hi @Huage001 , I make `inference_t2i.ipnyb` to .py file (code is exactly the same), and try to test the acceleration in flux dev, I meet the error in pipe inference...

Inference time saving for 1024 * 1024

1

Thanks for your excellent work on Flux, I see the flops saved in the table, can you share the inference time saving in GPU in resolution=1024 * 1024 ?