Commit Graph

294 Commits

Author SHA1 Message Date
filipstrand
d80b6e520f Add option to override saved images 2025-03-01 21:34:26 +01:00
filipstrand
8e5e09f7d8 Allow latest mlx 2025-02-19 08:09:34 +01:00
filipstrand
b733452878 Temporarily cap mlx at <0.23.0 until tests pass 2025-02-17 20:42:43 +01:00
Filip Strand
dca6c0acea
Merge pull request #126 from filipstrand/img2img-special-case
Img2img special case fix
2025-02-14 10:41:02 +01:00
filipstrand
821238284a Fix: Handle edge case for img2img 2025-02-14 05:21:35 +01:00
filipstrand
74173f543e Let before-loop callback accept latents and let StepwiseHandler include initial latent 2025-02-14 05:21:18 +01:00
Filip Strand
0417852a9b
Merge pull request #125 from filipstrand/more-quantization
More quantization options
2025-02-12 23:05:08 +01:00
filipstrand
3cecfae75b Refactor model config 2025-02-12 22:27:34 +01:00
filipstrand
f21710eb7e Update reference test images (minor changes due to updated quantization call) 2025-02-12 22:27:34 +01:00
filipstrand
4563bdd34b Update licence 2025-02-12 22:27:34 +01:00
filipstrand
a26277a7dd Require mlx >= v0.22.0 2025-02-12 22:27:34 +01:00
filipstrand
e3b2a65aa6 Allow 3-bit and 6-bit quantization (mlx > v0.21.0) 2025-02-12 22:27:34 +01:00
filipstrand
09d75e46b8 Add PromptCache: Small optimization for repeated generations with same prompt 2025-02-12 22:27:34 +01:00
Filip Strand
2d481e77b2
Merge pull request #124 from filipstrand/refactor-and-add-callback-mechanism
Refactor and add callback mechanism
2025-02-10 18:53:52 +01:00
filipstrand
3601baf75e Implement general callback mechanism 2025-02-09 15:27:06 +01:00
filipstrand
e4af40e4ee Refactor: Remove ConfigControlnet 2025-02-09 14:37:23 +01:00
filipstrand
22f98fc5a4 Refactor: Add separate Flux initializer 2025-02-09 14:37:23 +01:00
Filip Strand
8f82549f43
Merge pull request #122 from filipstrand/transformer-refactor
Transformer refactor
2025-02-02 19:13:45 +01:00
filipstrand
b5db14b3c6 Add descriptive comments for attention computations 2025-02-02 19:12:07 +01:00
filipstrand
0485cf07c2 Refactor: Transformer blocks 2025-02-02 19:12:07 +01:00
filipstrand
0d9a5d626f Refactor: Transformer & Flux 2025-02-02 19:12:07 +01:00
filipstrand
fa99cf0cc8 Refactor: Controlnet transformer & Flux controlnet 2025-02-02 16:45:27 +01:00
filipstrand
010610e33b Remove comments 2025-01-28 00:04:35 +01:00
Filip Strand
d34e60365d
Merge pull request #121 from filipstrand/training-on-non-sequential-block-layers
Training on non sequential block layers
2025-01-27 23:54:33 +01:00
filipstrand
f94edc4c3b Some fixes: Removed time in plotting, fix off-by-one error in training spec, update y-axis on loss plot 2025-01-27 23:48:03 +01:00
Alessandro
401da9fd75 Rebased on main (by filipstrand) 2025-01-27 23:15:27 +01:00
Filip Strand
cb93e1aaf6
Merge pull request #120 from anthonywu/renamer-uv-script-tool
Batch Image Renamer tool as an isolated uv run script
2025-01-27 22:23:18 +01:00
Anthony Wu
9053bcd6d8 Batch Image Renamer tool as an isolated uv run script 2025-01-22 13:25:59 -08:00
Filip Strand
4bea7b9668
Merge pull request #118 from anthonywu/run-n-seeds-auto-seeds
support multiple seeds batch generation
2025-01-22 19:58:51 +01:00
Filip Strand
fbf994a83e
Merge pull request #119 from filipstrand/support_variable_number_of_transformer_blocks
Support variable numbers of transformer blocks
2025-01-22 19:39:38 +01:00
filipstrand
807b243e56 Support variable numbers of transformer blocks 2025-01-22 08:18:39 +01:00
Anthony Wu
af01fab7d5 support multiple seeds batch generation 2025-01-21 18:14:24 -08:00
filipstrand
6fbbc9e5f7 Update missing model config lookup 2025-01-19 23:23:19 +01:00
Filip Strand
976573c837
Merge pull request #103 from anthonywu/support-hf-models
release 0.6.0 – comprehensive ModelConfig refactor to support compatible HuggingFace dev/schnell models
2025-01-19 23:13:37 +01:00
filipstrand
398cf082c3 Fix string formating 2025-01-19 19:49:57 +01:00
filipstrand
21a98b1369 Small updates 2025-01-19 15:19:17 +01:00
Anthony Wu
344ac90c74 comprehensive ModelConfig refactor to support compatible HuggingFace dev/schnell models 2025-01-17 06:27:33 -08:00
filipstrand
e32be7bf31 Update reference test images to match output (mlx=v0.22.0 and changed precision in attention mask)
Noticeable differences in images, but should be OK
2025-01-14 22:15:01 +01:00
filipstrand
16a0225b41 Specify precision for causal attention mask 2025-01-14 21:32:02 +01:00
Filip Strand
3b0f5a4516
Merge pull request #116 from filipstrand/optimizations
Optimizations and small refactoring of attention blocks
2025-01-14 21:18:04 +01:00
filipstrand
9954e7fb45 Unify attention mechanism in single and joint attention blocks 2025-01-14 20:16:48 +01:00
filipstrand
f0e82ad67f Optimization: Use fast SDPA (no speedup for me on tests, but a bit cleaner)
Had to update reference test images, but visually no difference
2025-01-14 20:16:23 +01:00
filipstrand
0eb8f5082e Update README with perf. update, installation issue and version link 2024-12-26 21:03:54 +01:00
filipstrand
2ebb0d1c6e Bump version to v.0.5.1 2024-12-23 12:40:19 +01:00
filipstrand
4c0aff9168 Fix bug with quantization level being str instead of int 2024-12-23 12:21:06 +01:00
Filip Strand
f725583bb9
Merge pull request #83 from filipstrand/dreambooth
DreamBooth support
2024-12-22 20:02:11 +01:00
filipstrand
bae49c8145 Dreambooth v1 2024-12-22 19:42:37 +01:00
filipstrand
627398f531 Add 'FLUX.1 Tools' for features to support 2024-11-21 21:31:27 +01:00
filipstrand
d69f49f31f Add In-Context LoRA for the cool research section 2024-11-20 00:18:29 +01:00
filipstrand
121a4f1695 Update reference test images to be compatible with mlx>0.19.0 (still visually identical) 2024-11-16 20:09:29 +01:00