Skip to content

Issues: kohya-ss/sd-scripts

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

687 Open 549 Closed
687 Open 549 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Issues list

train_single_block_indices Parameter Appears Ineffective
#1970 opened Mar 3, 2025 by Brackarnwar
1
Can I just save merged model after lora training for sd3.5-large?
#1969 opened Mar 3, 2025 by moonlightian
Why are model names hardcoded in analyze_checkpoint_state?
#1968 opened Mar 2, 2025 by CaledoniaProject
2
--cache_text_encoder_outputs_to_disk seems broken (in sdxl_train_control_net_lllite.py)
#1963 opened Mar 1, 2025 by Nekotekina
2
flux_train_network.py not support min_snr_gamma
#1962 opened Mar 1, 2025 by SlZeroth
'timestep_sampling=flux_shift' can offer quality improvements, but barely mentioned in docs
#1958 opened Feb 27, 2025 by araleza
1
Looking to understand this repos branching strategy
#1957 opened Feb 27, 2025 by davecranwell
controlnet训练
#1956 opened Feb 27, 2025 by lsx193201
SDXL seems to not train self_attn layers in Text Encoders
#1952 opened Feb 25, 2025 by Nekotekina
1
3
svd_merge_lora ver.flux
#1949 opened Feb 22, 2025 by leetherO
Should we have a "train_native.py"?
#1947 opened Feb 21, 2025 by 6DammK9
At resuming: RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!
#1944 opened Feb 18, 2025 by IAMAl
12
Why does the following error occur when running the latest flux_train_network.py with PyTorch 2.6.0+CUDA 12.4?
#1941 opened Feb 17, 2025 by zakinp
1
SDXL LoRA Training Stuck at Step 0
#1939 opened Feb 16, 2025 by hanpi2233
1
Bug❓: --cache_latents prevents regular checkpoint saving in train_network.py 😥
#1937 opened Feb 15, 2025 by Jackiiiii
2
WARNING Unsuppored keys for sd-scripts ！
#1933 opened Feb 14, 2025 by Shmtu-Herven
Train flux api
#1932 opened Feb 14, 2025 by iemskl
use multigpu(8*A800 80G) train flux_train.py OOM problem
#1930 opened Feb 13, 2025 by godwenbin
4
The lora+ setting and prodigy cannot be used together.
#1929 opened Feb 12, 2025 by ytoaa
1
Can the algorithm for automatic scaling of ARB buckets be specified?
#1928 opened Feb 12, 2025 by TLFZ1
1
Separate Learning Rates for UNET and TE with Prodigy
#1926 opened Feb 9, 2025 by MarvyD
1
Multi GPU need a fix : to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect.Specify device_ids in barrier() to force use of a particular device,or call
#1925 opened Feb 7, 2025 by FurkanGozukara
1
Support Lumina 2.0
#1924 opened Feb 7, 2025 by sdbds
11
Why have the options to define the single and double blocks_to_swap been removed?
#1923 opened Feb 7, 2025 by Deathawaits4
1
--save_state doesn't produce anything?
#1921 opened Feb 5, 2025 by alexgilseg
3
ProTip! Follow long discussions with comments:>50.