search

cocktailpeanut / fluxgym

Dead simple FLUX LoRA training UI with LOW VRAM support
1.13k stars 89 forks source link

How can I change the batch size? #134

Open MLGODFATHER opened 4 weeks ago

MLGODFATHER commented 4 weeks ago

How to Change Batch_Size?

When I try to change the batch size within the Train Config on the far-right column it just quickly reverts back to the default of 1. Can anyone tell me how I can change the batch size to make the traiging faster?

FNVDesigns commented 3 weeks ago

@MLGODFATHER You can change the batch size in the advanced settings. But when I change this, I get an error message. But maybe you have more luck.

MLGODFATHER commented 3 weeks ago

@MLGODFATHER You can change the batch size in the advanced settings. But when I change this, I get an error message. But maybe you have more luck.

I don't see that option anywhere...

InB4DevOps commented 2 weeks ago

@MLGODFATHER You can change the batch size in the advanced settings. But when I change this, I get an error message. But maybe you have more luck.

I don't see that option anywhere...

We can't help you to read but it looks like this:

image

Vigilence commented 2 weeks ago

@InB4DevOps Not sure if that option worked when you tried it, but its currently over ridden by the batch size hard coded in the python script to run flux gym. It will ignore what ever you put in there.

@MLGODFATHER @FNVDesigns See this for a solution.

MLGODFATHER commented 2 weeks ago

@InB4DevOps Not sure if that option worked when you tried it, but its currently over ridden by the batch size hard coded in the python script to run flux gym. It will ignore what ever you put in there.

@MLGODFATHER @FNVDesigns See this for a solution.

Thanks so much for the info!!

flashbackpoductions commented 1 week ago

@InB4DevOps Not sure if that option worked when you tried it, but its currently over ridden by the batch size hard coded in the python script to run flux gym. It will ignore what ever you put in there. @MLGODFATHER @FNVDesigns See this for a solution.

Thanks so much for the info!!

hi - i change the app.py code manually in the gen toml function 👍

def gen_toml( dataset_folder, resolution, class_tokens, num_repeats ): toml = f"""[general] shuffle_caption = false caption_extension = '.txt' keep_tokens = 1

[[datasets]] resolution = {resolution} batch_size = 2 keep_tokens = 1

[[datasets.subsets]] image_dir = '{resolve_path_without_quotes(dataset_folder)}' class_tokens = '{class_tokens}' num_repeats = {num_repeats}""" return toml

i will report if there is a change