sd-scripts
sd-scripts copied to clipboard
kohya-ss
The Accelerator setup / etc was using loop, with explicit local process index check instead of process index check, resulting multi-node training hang forever. After struggling against code for weeks,...
Question: How do SD's text_tokenizer and Unet work when the input prompt is too long? Description: Hello, esteemed expert! I have a question recently. When I use AUTOMATIC1111/stable-diffusion-webui, I found...
Subject: Proposal for Implementing a 1-bit Optimizer Dear [sd-scripts] Team, I hope this message finds you well. I am reaching out to propose the addition of a 1-bit optimizer to...
If you are using a custom amount of tokens rather then the default in clip, then it errors out with a size mismatch. Is there a reason to set the...
hello, there exists a parameter to pass a wandb token into the training utility. unfortunately, this shows on the "Overview" tab of the training session, as a part of the...
As the title says, we are still experiencing very slow perfomance on rtx 4090 cards(and probably some other cards to) we would really like to find a solution to this,...
This change adds the parameters used with `accelerate` and sends them as a payload to `wandb.ai`. That allows the parameters to show up in the table of runs and helps...
At line 1711 of `sdxl_gen_img.py`, the script calls the `is_mergeable()` method to check whether a network is mergeable. But this causes an error when using dylora, since the `DyLoRANetwork` class...
Hello, I used the changes made by @larsupb, that were mistakenly proposed in @bmaltais's repository, with some adjustments I did to fit the current state of `train_network.py`, to enable support...
I trained a lora with kohya_ss and Onetrainer, I'm confused that why the lora with kohya_ss is much faster than using diffusers? If the network is different from others, why...
