ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
963 commits 391 branches 58 tags 304 MiB
Commit graph

963 commits

This branch
This branch
All branches
Author SHA1 Message Date
Roland Tannous
2cfcaa8b61 feat: broaden Qwen3.5 matching to cover entire family 2026-03-06 16:48:28 +00:00
Roland Tannous
c31e8a6ed7 feat: add Qwen3.5-35B-A3B and Qwen3-Next to transformers 5.x model list 2026-03-06 10:54:48 +00:00
Manan17
8afaf55eb5 derive effective model type from isVisionModel for dataset search filterin 2026-03-06 08:47:56 +00:00
Shine1i
72468a3292 feat(recipe-studio, validators): tweak OXC validator with lint suppression support and improve error normalization logic 2026-03-06 09:40:53 +01:00
Roland Tannous
527db4ffff feat: add OpenAI-compatible /v1/chat/completions endpoint 2026-03-06 07:48:09 +00:00
Manan17
fcd52d9bd0 fixing update model type 2026-03-06 07:44:57 +00:00
imagineer99
e0e4619ee2 fix: harden chart container sizing with legacy event rechecks 2026-03-06 07:16:36 +00:00
Roland Tannous
7e59440029 fix: pin huggingface_hub==1.3.0 in .venv_t5 (satisfies transformers 5.x) 2026-03-06 06:19:28 +00:00
imagineer99
e231f366ea fix: preserve chart sizing updates without ResizeObserver 2026-03-06 06:06:50 +00:00
Roland Tannous
c82e3d86bb fix: unload competing subprocesses before load across all routes 2026-03-06 06:05:31 +00:00
Roland Tannous
661ac4be96 feat: subprocess-based export, pin huggingface_hub==0.36.0 2026-03-06 06:03:09 +00:00
imagineer99
6f13185379 fix: align slider fill bar with thumb across value range 2026-03-06 05:21:44 +00:00
Roland Tannous
f15970c02a fix: use subprocess with transformers 5.x for vision detection 
Models like GLM-4.7-Flash have architectures (glm4_moe_lite) that
AutoConfig in the main process (transformers 4.57.x) can't recognize.
Instead of a raw config.json workaround, run the AutoConfig check in
a subprocess with .venv_t5/ activated — same pattern as training and
inference workers. This is more robust and consistent.
 2026-03-06 04:51:23 +00:00
Roland Tannous
67121ce427 fix: handle unrecognized model architectures in vision detection 
AutoConfig.from_pretrained() fails for models needing transformers 5.x
(e.g. glm4_moe_lite) when running with 4.57.x. Add a raw config.json
fallback that bypasses AutoConfig's architecture registry — fetches
config.json directly from local path or HuggingFace Hub and checks
for vision indicators without needing the architecture to be registered.
 2026-03-06 04:46:51 +00:00
Roland Tannous
d55e9abcca refactor: consolidate version switching to .venv_t5, remove .venv_overlay 
All version switching now uses .venv_t5/ (pre-installed by setup.sh).
The old .venv_overlay/ with runtime pip installs is removed.
ensure_transformers_version() (used only by export) now does a
lightweight sys.path swap instead of pip installing at runtime.
 2026-03-06 04:37:06 +00:00
Manan17
cf9eaa2add Fixing dataset split issues 2026-03-06 01:13:55 +00:00
Shine1i
4bd7eca3f3 feat(recipe-studio, validators): extend OXC validator with code shape support and integrate into recipe studio 2026-03-06 02:04:05 +01:00
Shine1i
769515cb5f feat(recipe-studio): add inference_timeout configuration and validation logic 2026-03-06 00:51:28 +01:00
Shine1i
5c33886cce feat(recipe-studio): add support for inference_extra_body configuration with collapsible UI and enhanced validation logic 2026-03-05 23:33:48 +01:00
Roland Tannous
ce9bfd7476 fix: unload inference model before training to free GPU memory 
When starting training, shut down the inference subprocess first
so the training subprocess has full GPU memory available.
 2026-03-05 22:28:11 +00:00
imagineer99
4c634c4d2a fix: guard recharts ResponsiveContainer behind measured container dimensions 2026-03-05 21:54:27 +00:00
Roland Tannous
e866d16787 fix: indentation error in orchestrator load_model 2026-03-05 19:43:30 +00:00
Shine1i
d687ca6437 feat(data-recipes, validators): extend OXC validator with linting mode support and integrate new modes into recipe studio 2026-03-05 20:19:31 +01:00
Roland Tannous
b598cf0c14 fix: always spawn fresh subprocess per model load 
Reusing a subprocess after unsloth patches torch internals causes
inspect.getsource() failures when loading a different model type.
Each load now gets a clean Python interpreter.
 2026-03-05 19:15:37 +00:00
Roland Tannous
5c3efd610f fix: use mp.Event for instant cross-process generation cancel 
Replaces cmd_queue-based cancel polling with a shared mp.Event.
Fixes two issues:
- Loading a new model while generating no longer hangs (cancel is instant)
- Subprocess shuts down cleanly after explicit stop generation
 2026-03-05 18:54:17 +00:00
Shine1i
f9eabf285d feat(data-recipes, validators): add OXC validator runtime and integration with recipe studio 2026-03-05 19:48:26 +01:00
Roland Tannous
b70faf8cb7 feat: subprocess-based inference for transformers version switching 
Inference now runs in a persistent subprocess, solving the same
transformers version-switching problem that was fixed for training.
The subprocess stays alive between requests (model in GPU memory)
and is only restarted when switching transformers versions.

New files:
- core/inference/worker.py: subprocess entry point with command loop
- core/inference/orchestrator.py: parent-side proxy with same API

Modified:
- core/inference/__init__.py: exports orchestrator as default backend
- routes/inference.py: removed in-process ensure_transformers_version()
 2026-03-05 17:47:57 +00:00
Roland Tannous
021c3aafdd fix: handle None job_id before first training run 2026-03-05 16:59:37 +00:00
Roland Tannous
36b5c6af88 fix: lazy imports in core/__init__ to prevent subprocess importing ML libs early 2026-03-05 16:56:45 +00:00
Roland Tannous
794b8fe866 fix: exclude bitsandbytes from module purge to prevent duplicate operator registration 2026-03-05 16:40:20 +00:00
Roland Tannous
06fbeb22cf fix: remove in-process version switching from models routes 2026-03-05 16:22:32 +00:00
Roland Tannous
fb29d7f999 fix: remove UnslothTrainer/get_trainer from core __init__ exports 2026-03-05 15:57:07 +00:00
Roland Tannous
f90af41c5f feat: subprocess-based training for transformers version switching 2026-03-05 15:40:32 +00:00
Shine1i
6fc829e3ef merge: nightly into feature/data-reciper-enchansments 2026-03-05 14:51:08 +01:00
Shine1i
3ff0e74216 feat(recipe-studio): runtime edge handling with template refs and reversed edge support 2026-03-05 14:46:48 +01:00
Shine1i
b54238bbf5 refactor(studio): replace inputValue with searchQuery for improved clarity, add input reason tracking, and streamline dataset filtering logic 2026-03-05 14:06:17 +01:00
Shine1i
5674c6a815 feat(recipe-studio): improve tab switch fit logic with animation and delay support 2026-03-05 13:29:38 +01:00
Shine1i
2ff1d4de36 feat(recipe-studio): normalize and slugify run_name, update job naming logic 2026-03-05 12:25:51 +01:00
Shine1i
f988202290 refactor(studio): add local data-recipe dataset selection + training wiring 2026-03-05 12:25:51 +01:00
Shine1i
59995d8447 feat(data-recipes, recipe-studio): refactor and enhance recipe templates with updated model configurations, structure changes, and added validation logic 2026-03-05 12:14:01 +01:00
Shine1i
015e93491f feat(recipe-studio): persist advanced collapsible states across components and sessions 2026-03-05 11:56:40 +01:00
Shine1i
e50b2345fd feat(data-recipes, recipe-studio): recipies changes, image context selector 2026-03-05 11:46:42 +01:00
Shine1i
b43e9033fd feat(data-recipes): update recipe templates 2026-03-05 11:29:20 +01:00
Manan17
5d1a162ddd remove tracked OuteTTS embedded repo reference 2026-03-05 08:44:23 +00:00
Manan17
8203637d89 resolved merge conflicts 2026-03-05 07:59:43 +00:00
Manan17
2fa933640e fix SNAC training crash on variable-length sequences with DataCollatorForSeq2Seq 2026-03-05 07:04:53 +00:00
Roland Tannous
f57664e268 Merge nightly into feature/transformers-v5-support 2026-03-05 06:49:44 +00:00
Roland Tannous
37bff450d5 Merge pull request #314 from unslothai/fix/vlm-dataset-conversion-error-handling-local 
Fix VLM training abort on URL-based dataset conversion failure
 2026-03-05 10:10:58 +04:00
Roland Tannous
a1706c894f fix: check for http(s) prefix instead of bare string type for URL detection 2026-03-05 06:10:10 +00:00
Roland Tannous
2116cc5cca fix: remove benchmark scripts from git tracking 
These are standalone benchmark scripts that were force-added despite being
gitignored. They have no test functions and run network calls at module
level, which breaks pytest collection in CI.
 2026-03-05 06:06:47 +00:00