Skip to content

Pull requests: kohya-ss/sd-scripts

New pull request New
114 Open 595 Closed
114 Open 595 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
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Fix/leco cleanup
#2294 opened Mar 28, 2026 by kohya-ss Loading…
feat: Add support for Safetensors format in caching strategies
#2290 opened Mar 22, 2026 by kohya-ss • Draft
Log d*eff_lr for ProdigyPlusScheduleFree, and some cleanup of LR logging
#2289 opened Mar 21, 2026 by woct0rdho Loading…
Use torch.svd_lowrank for large matrices in resize_lora.py
#2240 opened Nov 17, 2025 by woct0rdho Loading…
1
Make stable validation loss stable and deterministic for Flux when using uncached latents
#2238 opened Nov 12, 2025 by mturnshek Loading…
1
Add fp8 scaling, fp8 scaled convolutions, and gradient checkpointing for Flux AutoEncoder
#2231 opened Oct 29, 2025 by rockerBOO • Draft
Add BREAK for captions in strategy_sd
#2223 opened Oct 10, 2025 by rockerBOO • Draft
Add non_blocking to loading and moving tensors
#2222 opened Oct 10, 2025 by rockerBOO • Draft
1
CDC-FM (Carré du Champ Flow Matching) Implementation
#2221 opened Oct 9, 2025 by rockerBOO Loading…
6
feat: torch compile within checkpointing and activation memory budget for Lumina 2
#2217 opened Sep 30, 2025 by urlesistiana Loading…
7
Handle icc profiles and images with transparency
#2210 opened Sep 27, 2025 by Hirmuolio Loading…
1
7
feat: add fp8 optimization for FLUX
#2208 opened Sep 25, 2025 by kohya-ss • Draft
2
Chroma fix t5
#2203 opened Sep 20, 2025 by johnr14 • Draft
5
Fix the 'off by 1' problem in dynamically resized LoRA rank
#2194 opened Sep 4, 2025 by woct0rdho Loading…
fix: Fix variable reference error in sd-scripts/library/flux_models.py on sd3 branch.
#2188 opened Aug 25, 2025 by Holasyb918 Loading…
2
Amazing Flux FFT quality on a 5090: (N)AdamW / Adan with fused backwards pass & CPU offloading
#2187 opened Aug 24, 2025 by araleza Loading…
7
Add MGPO to Flux network
#2180 opened Aug 19, 2025 by rockerBOO • Draft
2
Add support for DoRA and diffusions layer names
#2172 opened Aug 10, 2025 by Reithan Loading…
Big image quality improvement! Kahan summation for Adafactor-optimized Flux FFT
#2159 opened Jul 23, 2025 by araleza Loading…
20
Added warning for buckets without enough images to half fill a batch
#2141 opened Jul 1, 2025 by araleza Loading…
3
train_network.py: add debugging print for tensor and gradient norms
#2136 opened Jun 28, 2025 by ddpasa Loading…
3
Add IVON optimizer support
#2129 opened Jun 19, 2025 by rockerBOO • Draft
Update tag_images_by_wd14_tagger.py
#2109 opened Jun 4, 2025 by matan2244 Loading…
Support Cogview4
#2097 opened May 26, 2025 by sdbds • Draft
Randomized and Incremental Regularization Image Loading - Overhaul of Regularization Image Loading Logic
#2096 opened May 25, 2025 by DKnight54 Loading…
1
ProTip! Adding no:label will show everything without a label.