ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
1,060 commits 391 branches 58 tags 304 MiB
Commit graph

1,060 commits

This branch
This branch
All branches
Author SHA1 Message Date
Roland Tannous
e455b307be add fmpeg system support for linux and windows 2026-03-11 18:50:11 +00:00
Roland Tannous
a6aa9a0efa Remove test_llama_cpp.ps1 from tracking 2026-03-11 18:10:25 +00:00
Roland Tannous
859bfe23c4 Merge pull request #375 from unslothai/feature/llm-assist-detection 
Feature/llm assist detection
 2026-03-11 22:02:33 +04:00
Roland Tannous
b274e9e0c6 chore: merge nightly & update dataset preview dialog mapping text 2026-03-11 17:00:14 +00:00
Roland Tannous
0e3ac91e2a feat: target AI Assist mapping prompts for audio & embedding models 2026-03-11 16:55:43 +00:00
Roland Tannous
9dac1bedf9 Merge remote-tracking branch 'origin/nightly' into feature/llm-assist-detection 2026-03-11 16:23:09 +00:00
Roland Tannous
6d6a62821e Merge pull request #374 from unslothai/fix/model-caching-issues 
Fix: Normalize HuggingFace model identifiers to lowercase
 2026-03-11 18:40:26 +04:00
Roland Tannous
7862e70211 fix: lowercase remote Hugging Face model IDs in ModelConfig and routes to prevent caching mismatches with Unsloth 2026-03-11 14:20:25 +00:00
Roland Tannous
fb211f3254 Merge pull request #372 from unslothai/fix/input-focus-clipping 
Input focus outline clipped by container
 2026-03-11 18:12:04 +04:00
Roland Tannous
774c9b17fd Merge pull request #373 from unslothai/feature/structlog-logging-system 
feat: integrate structlog, configure workers for prod logging, and mi…
 2026-03-11 16:52:49 +04:00
Roland Tannous
817f2e8dcc feat: integrate structlog, configure workers for prod logging, and migrate print statements 2026-03-11 12:33:16 +00:00
imagineer99
014695b38a fix: scope overflow-visible to studio collapsibles 2026-03-11 11:26:43 +00:00
imagineer99
984f4a4acb fix: input focus outline clipping 2026-03-11 11:11:57 +00:00
Roland Tannous
ee063c5910 Merge pull request #367 from unslothai/fix/yaml-syntax 
Modified to fix the yaml syntax for unsloth_Qwen3-14B-Base-unsloth-bnb-4bit
 2026-03-11 13:39:48 +04:00
Roland Tannous
16eb39b53d Merge pull request #369 from unslothai/fix/model-mappping-syntax 
fixed string concatenation in model mapping
 2026-03-11 12:39:13 +04:00
Samit
379bbbdbdd fixed string concatenation in model mapping 2026-03-11 00:07:26 -07:00
Samit
822050bf57 modified to fix the yaml syntax 2026-03-10 23:58:51 -07:00
Manan Shah
2aa9322167 Merge pull request #365 from unslothai/fix/gguf-gemma-with-text 
fixing gguf export for gemma with text
 2026-03-10 17:59:22 -07:00
Manan17
5ca623a166 fixing gguf export for gemma with text 2026-03-11 00:58:22 +00:00
Wasim Yousef Said
739838bd48 Merge pull request #364 from unslothai/feature/chat-seq-slider 
chat seq slider
 2026-03-11 01:56:48 +01:00
Shine1i
4a8a96b1af chat seq slider 2026-03-11 01:41:25 +01:00
Manan Shah
cce274717b Merge pull request #357 from unslothai/feat/embedding-models 
feat: add embedding model training support
 2026-03-10 14:59:20 -07:00
Manan17
983c20bbb2 local model's embedding nature check 2026-03-10 21:58:45 +00:00
Manan17
294a3d3e47 fix: reset isEmbeddingModel in error fallback paths to prevent stale state 2026-03-10 21:33:13 +00:00
Roland Tannous
5b042165e6 Merge pull request #363 from unslothai/feature/enable-all-modalities 
Removed audio and embedding from coming soon
 2026-03-11 01:32:43 +04:00
Manan17
bc5a72dd8c fix: local directory dataset loading 2026-03-10 21:29:51 +00:00
imagineer99
8d6f88577f chore: removed audio and embedding from coming soon 2026-03-10 21:29:18 +00:00
Wasim Yousef Said
98e3396fbe Merge pull request #356 from unslothai/fix/summary-step-spacing-and-colors 
Redesign summary step with consistent card layout, spacing and icons
 2026-03-10 22:26:10 +01:00
Wasim Yousef Said
27311e3986 Merge pull request #362 from unslothai/feature/setup-no-llama-nuke 
fix(setup): stop deleting llama.cpp in setup
 2026-03-10 22:25:12 +01:00
Manan Shah
f696ef81e8 Merge branch 'nightly' into feat/embedding-models 2026-03-10 14:16:05 -07:00
Roland Tannous
08d9c84f1f Merge pull request #359 from unslothai/fix/stream-manual-slice-dataset 
fix: stream HF dataset when manual slice is specified
 2026-03-11 01:13:51 +04:00
Manan17
9523e5c1f9 fixing embedding model search 2026-03-10 21:12:24 +00:00
Shine1i
2895518f0c fix(setup): stop nuking llama.cpp in setup 2026-03-10 22:03:01 +01:00
Wasim Yousef Said
29e56e8649 Merge pull request #361 from unslothai/fix/tooltip-z-index 
Increase tooltip z-index to appear above dropdowns
 2026-03-10 22:01:30 +01:00
imagineer99
d572c43814 fix: increase tooltip z-index to appear above dropdowns 2026-03-10 20:57:12 +00:00
Manan17
3b0b002b34 fixing logging for each step 2026-03-10 20:32:40 +00:00
Roland Tannous
21ef22a9ff fix: skip streaming when dataset_slice_start > dataset_slice_end 
Prevents training on the wrong row range when start exceeds end by
falling back to full download where existing clamping handles it.
 2026-03-10 20:21:34 +00:00
imagineer99
5dcbf86d09 fix: reject negative manual dataset slices 
Prevent negative Train Split Start/End values in the dataset advanced UI and sanitize payload mapping so negative slice values are never sent to the backend.

Made-with: Cursor
 2026-03-10 20:13:46 +00:00
Roland Tannous
226f251589 fix: guard against negative dataset_slice_end before streaming 
Fall back to full download when dataset_slice_end is negative,
avoiding an empty stream.take(0) that would produce a broken dataset.
 2026-03-10 20:12:42 +00:00
Roland Tannous
b91cdda2b9 Merge pull request #354 from unslothai/fix/audio-train-completions 
fix: uncheck train_on_completions for audio models
 2026-03-11 00:05:51 +04:00
Roland Tannous
949f2ac87e Merge pull request #358 from unslothai/fix/sharded-gguf 
fix: download all GGUF shards for split models
 2026-03-11 00:05:01 +04:00
Roland Tannous
970a029108 fix: stream HF dataset when manual slice is specified 
Instead of downloading the full dataset and then slicing, use
streaming mode to only fetch the rows needed (up to slice_end + 1)
when a manual dataset slice is configured.
 2026-03-10 19:50:53 +00:00
Roland Tannous
c986174c56 fix: preserve zero-valued dataset slice boundaries in embedding worker 
Use explicit None checks instead of falsy `or` for slice_start and
slice_end so that a valid slice_end=0 is not replaced with the full
dataset length.
 2026-03-10 19:33:10 +00:00
Roland Tannous
b84202e8db fix: restrict shard siblings to exact basename and total count 
startswith(prefix) could match unrelated split variants whose names
extend the selected file's prefix (e.g. model-Q8_0-v2-00001-of-...).
Now builds an exact regex from the chosen file's base prefix and shard
total so only true siblings are downloaded.
 2026-03-10 19:28:26 +00:00
Shine1i
18a60b930a chore/fix(studio): add placeholder dropdowns for dataset subset and splits in disabled state 2026-03-10 20:27:11 +01:00
Roland Tannous
b8678a3ed6 fix: pass hf_token for gated embedding models and key cache by token 
- Forward hf_token to FastSentenceTransformer.from_pretrained() so
  private/gated embedding repos authenticate correctly
- Key _embedding_detection_cache by (model_name, hf_token) tuple so
  unauthenticated lookups don't shadow subsequent authenticated ones
 2026-03-10 19:20:12 +00:00
Roland Tannous
d635846b8d fix: use exact variant matching and shard-prefix discovery for split GGUFs 
Substring matching (e.g. "Q8_0" in filename) could match superset
variants like "IQ8_0", causing wrong quantizations to be downloaded.
Now uses word-boundary regex for variant matching and discovers split
shards by shared filename prefix rather than treating all variant
matches as shards.
 2026-03-10 19:13:03 +00:00
Roland Tannous
d6ae910edc fix: propagate is_embedding into worker subprocess config 
start_training() cherry-picks kwargs into a config dict but was missing
is_embedding, so config.get("is_embedding", False) in worker.py always
returned False and embedding training never ran.
 2026-03-10 19:05:47 +00:00
Roland Tannous
defa761fb2 fix: download all GGUF shards for split models (e.g. 7B Q8_0) 
LlamaCppBackend.load_model() only downloaded the first matching GGUF
file. For split models (e.g. 7B Q8_0 with 3 shards), llama-server
needs all shards present. Now collects and downloads all matching files.
 2026-03-10 19:04:10 +00:00
Roland Tannous
846cc2cf2a fix: always force-uncheck trainOnCompletions for pure audio models in dataset check 
Separate pure-audio from audio-VLM logic in runDatasetCheck so pure
audio models are always forced to trainOnCompletions=false regardless
of dataset type, while audio VLMs (gemma3n) only uncheck when the
dataset is audio.
 2026-03-10 19:02:49 +00:00