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

112 commits

Author SHA1 Message Date
Shine1i
acf7cce4a8 merge: nightly into feature/data-reciper-enchansments 
resolve setup.sh conflict by keeping nightly installer flow and preserving local data-designer plugin install via install_python_stack.py
 2026-03-03 22:21:04 +01:00
Roland Tannous
5f98d232d0 fix: align llama-server binary discovery with upstream unsloth-zoo paths 2026-03-03 17:03:01 +00:00
Shine1i
e2292df7da refactor(seed): package unstructured seed reader as local Data Designer plugin 2026-03-03 11:22:04 +01:00
Shine1i
a2b61fa762 merge nightly into feature/data-reciper-enchansments 2026-03-03 11:14:18 +01:00
Shine1i
4c5a2543c3 feat(seed): backend unstructured seed reader + server-side chunking, remove client chunk splitter 2026-03-03 11:11:26 +01:00
Roland Tannous
e7619a1291 Move llama.cpp clone/build from in-tree to ~/.unsloth/llama.cpp 
- setup.sh: builds at ~/.unsloth/llama.cpp instead of ./llama.cpp
- setup.ps1: builds at %USERPROFILE%/.unsloth/llama.cpp
- inference llama_cpp.py: searches ~/.unsloth/ first, in-tree as legacy
- export.py: updated comments (unsloth-zoo handles path natively)
 2026-03-02 04:04:41 +00:00
Roland Tannous
70d1567fe3 Download GGUF via huggingface_hub instead of llama-server -hf (fixes HTTPS not supported on Windows) 2026-03-01 13:05:10 +00:00
Roland Tannous
0430d22cc2 Auto-add CUDA DLLs to PATH when launching llama-server on Windows 2026-03-01 13:05:10 +00:00
Roland Tannous
afa1344452 Build llama.cpp in-tree, auto-detect driver CUDA version for compatible toolkit 2026-03-01 13:05:10 +00:00
Roland Tannous
af90c9c3d2 Fix llama-server binary lookup for Windows (.exe, Release dir, ~/.unsloth) 2026-03-01 13:05:10 +00:00
Roland Tannous
986bef4f99 fix: support mmproj for local vision GGUF models + fix Windows pipe deadlock 2026-03-01 12:58:38 +00:00
Shine1i
899350d3e8 feat(recipe-studio): add LLM trace modes and reasoning content extraction support 2026-03-01 12:01:48 +01:00
Shine1i
653b85a07e refactor(recipe-studio): add image preview support for dataset and LLM configurations p2 2026-03-01 11:21:10 +01:00
Shine1i
ac1c43ec77 feat(recipe-studio): add image preview support for dataset and LLM configurations p1 2026-03-01 10:57:51 +01:00
Manan17
b4311cca82 Aggregating sharded models, showing fit/oom for quantizations 2026-02-27 08:23:15 +00:00
Shine1i
2b2e870862 refactor(recipe-studio): simplify runtime graph flow + guard stale active execution lock p2 2026-02-26 15:37:48 +01:00
Shine1i
d356ded63a feat(recipe-studio): add live execution graph state (active flows, node status, editor lock) p1 2026-02-26 15:27:46 +01:00
Shine1i
46046cd118 refactor(data-recipe): centralize json+stage constants, tighten parser/errors, sync seed ui 2026-02-26 14:06:53 +01:00
Roland Tannous
c06adc3878 Flatten GGUF subdirs in export and fix metadata lookup in scanner 2026-02-26 11:35:04 +04:00
Roland Tannous
a715bca52f Write export metadata for GGUF exports to fix Unknown base model 2026-02-26 11:24:32 +04:00
Roland Tannous
f0176c7f0c Relocate GGUF exports into exports/ directory 2026-02-25 18:54:39 +04:00
Roland Tannous
efaa0bacfb Merge branch 'nightly' into feat/gguf-llama-cpp-inference 2026-02-25 16:06:03 +04:00
Roland Tannous
52738383f9 Remove UNSLOTH_ENABLE_LOGGING from export pipeline 2026-02-25 16:00:24 +04:00
Roland Tannous
d9434fee4a fix: use raw github URL for vision.py patch + add VLM processor diagnostic logging 2026-02-25 10:29:05 +00:00
Shine1i
e52e8a6d51 fix recipe output path, remove tracked root datasets 2026-02-25 11:19:10 +01:00
Roland Tannous
9e280eb105 Fix GGUF export cwd confusion: remove os.chdir, use absolute paths 
Remove os.chdir(save_directory) from export.py which was causing all of
unsloth-zoo's relative-path internals (check_llama_cpp, use_local_gguf,
_download_convert_hf_to_gguf) to resolve against the export directory
instead of the repo root. This caused llama.cpp to be cloned inside each
export dir and destroyed the repo root's llama-server build on cleanup.

Now passes absolute paths to save_pretrained_gguf so unsloth resolves
llama.cpp from the repo root where setup.sh already built it.

Also builds llama-quantize in setup.sh (needed by unsloth-zoo's export
pipeline) and symlinks it to llama.cpp root for check_llama_cpp().
 2026-02-25 03:30:54 +04:00
Roland Tannous
2ebeba8588 Switch GGUF backend from /v1/completions to /v1/chat/completions 
Fixes two bugs:
1. Chat template tags (<|im_start|>, <|im_end|>) leaking into output
   because /v1/completions treated them as literal text
2. Image hallucination because image_b64 was never passed to llama-server

Now llama-server handles chat templates natively and receives images
as OpenAI-format multimodal content parts for vision models.
 2026-02-24 19:21:01 +04:00
Roland Tannous
3ee4f1359a Use llama-server -hf mode, add GGUF variant selector, fix vision detection 
Replace Python-side GGUF download with llama-server's native -hf flag for
HuggingFace repos. Add frontend variant picker so users can choose
quantization (Q4_K_M, Q8_0, BF16, etc.) with file sizes. Fix vision
detection via mmproj files instead of hardcoding is_vision=False.
 2026-02-24 19:03:06 +04:00
Roland Tannous
5b7555cd3f Fix llama-server: build in-tree, fix path resolution, add LD_LIBRARY_PATH 2026-02-24 18:19:29 +04:00
Roland Tannous
2f985ccbb5 Add GGUF model inference via llama-server backend 2026-02-24 17:40:05 +04:00
Roland Tannous
a09823c2eb Merge branch 'nightly' into feature/canvas-lab 2026-02-24 10:08:13 +00:00
Roland Tannous
f8486672f6 Merge pull request #241 from unslothai/feature/adding-exported-models-for-chat 
Adding exported model for chat
 2026-02-24 13:55:45 +04:00
Roland Tannous
a661784d41 Merge pull request #232 from unslothai/fix/disable-eval-by-default 
# fix/disable eval by default
 2026-02-24 13:35:11 +04:00
Roland Tannous
a15082acfe skip eval split and HF split detection when eval_steps is disabled 2026-02-24 09:26:54 +00:00
Manan17
6fa6b0bf28 Fixing base model export issue for vlms 2026-02-24 01:34:11 +00:00
Manan17
1071c137f4 Adding exported model for chat 2026-02-24 01:17:09 +00:00
Shine1i
8ea9ac6a21 feat: add per-column seed drop support with UI integration, validation, and payload enhancements 2026-02-23 23:33:59 +01:00
Shine1i
ad097e359f Merge branch 'nightly' into feature/canvas-lab 2026-02-23 21:54:35 +01:00
Shine1i
5937ed76a2 refactor: add batch processing support with configuration options and execution enhancements 2026-02-23 21:32:20 +01:00
Shine1i
e61aab37d7 refactor: improve dialog rendering and logging setup for stability and configurability 2026-02-23 20:16:03 +01:00
Leo Borcherding
86388a0242 fix: disable eval by default, set eval_steps to 0.0 
- Changed default eval_steps from 0.01 to 0.0 across backend and frontend
- Fixed UI to allow eval_steps=0 (removed min=0.001 constraint)
- Added conditional eval logic with helpful console messages
- Updated tooltip to explain how to disable evaluation
- Tested: confirmed eval disabled by default with eval_steps=0.0
 2026-02-23 13:07:47 -06:00
Roland Tannous
834013aae5 Cap dataset.map num_proc on multi-GPU machines to prevent fork deadlocks 2026-02-23 14:25:31 +00:00
Roland Tannous
d94f842158 fix: error on >30% sample drop after train_on_responses_only instead of silent DataLoader crash 2026-02-23 12:21:06 +00:00
Roland Tannous
198433363a feat: clear unsloth_compiled_cache on startup, shutdown, and between model loads 2026-02-23 07:26:22 +00:00
Roland Tannous
e666442b6e fix: pass full Processor as processing_class for VLM SFTTrainer 2026-02-22 14:11:12 +00:00
Shine1i
66cceb69fb merge: nightly into feature/canvas-lab 2026-02-22 02:31:32 +01:00
Roland Tannous
ef118d0d05 fix: load proper vision processor from base model when FastVisionModel returns raw tokenizer, add tokenize=False to vision chat template 2026-02-21 04:40:29 +00:00
Manan17
e9710874e1 Mapping proper tokenizer for VLMs 2026-02-21 01:57:05 +00:00
Manan17
756aa56cd2 fixed the vlm's text only errors 2026-02-20 22:23:26 +00:00
Shine1i
394dc24048 feat: enhance execution log tracking, progress updates, and data visualization 
- Added `log_lines` field to track and display runtime logs for executions.
- Enhanced progress tracking with terminal-like log outputs and live log scrolling.
- Introduced detailed "model usage" and "dropped columns" analysis in `ExecutionsView`.
- Optimized UI components for displaying dataset metrics, including input/output token averages.
 2026-02-20 13:51:19 +01:00