VictorMylle/Kohya-ss-sd-scripts
1
0
Fork 1
mirror of https://github.com/kohya-ss/sd-scripts.git synced 2026-04-09 06:45:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

update readme

Browse Source
This commit is contained in:
Kohya S
2024-08-18 16:56:50 +09:00
parent ef535ec6bb
commit a450488928
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
README.md
View File

@@ -9,10 +9,8 @@ __Please update PyTorch to 2.4.0. We have tested with `torch==2.4.0` and `torchv
The command to install PyTorch is as follows:
`pip3 install torch==2.4.0 torchvision==0.19.0 --index-url https://download.pytorch.org/whl/cu124`
Aug 18, 2024:
Memory-efficient training based on 2kpr's implementation is implemented in `flux_train.py`. Thanks to 2kpr. See [FLUX.1 fine-tuning](#flux1-fine-tuning) for details.
Memory-efficient training based on 2kpr's implementation is implemented in `flux_train.py`. Thanks to 2kpr! See [FLUX.1 fine-tuning](#flux1-fine-tuning) for details.
Aug 17, 2024:
Added a script `flux_train.py` to train FLUX.1. The script is experimental and not an optimized version. It needs >28GB VRAM for training.
@@ -118,6 +116,8 @@ accelerate launch --mixed_precision bf16 --num_cpu_threads_per_process 1 flux_t
(Combine the command into one line.)
Sample image generation during training is not tested yet.
Options are almost the same as LoRA training. The difference is `--blockwise_fused_optimizer`, `--double_blocks_to_swap` and `--cpu_offload_checkpointing`. `--single_blocks_to_swap` is also available.
`--blockwise_fused_optimizer` enables the fusing of the optimizer for each block. This is similar to `--fused_backward_pass`. Any optimizer can be used, but Adafactor is recommended for memory efficiency. `--fused_optimizer_groups` is deprecated due to the addition of this option for FLUX.1 training.