Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.
Python agentaichatgptdeepseekfine-tuninggemmaimage-generationllamallmllmsopenaipythonqwenreinforcement-learningself-hostedstable-diffusiontext-to-speechttsuiunsloth
Commits on main
AMD RDNA1: turn Triton's buffer ops off on gfx101x before Triton loads (#11615)
41c0d44 Leo Borcherding committed 2 months ago
Read plain RoPE as rope_scaling None for 4.x-era remote configs on transformers 5 (#11611)
0065fad Daniel Han committed 2 months ago
Give each spawn-probe build its own package in the shared modules cache (#11672)
2ab56e5 Daniel Han committed 2 months ago
Studio: scale padding, icons and widths with the UI font size, and make interface scale work in the browser (#11648)
6eb1b2e Michael Han committed 2 months ago
Give the descendant-work health test room for its second interpreter to start (#11670)
beab871 Daniel Han committed 2 months ago
Fix Code tool placement for ChatGPT subscriptions (#11628)
8d632de lorenzozane committed 2 months ago
Studio: list the unsloth mirrors for curated BF16 image pipelines (#11657)
54964b1 Michael Han committed 2 months ago
Pin the last 53 mutable action tags, and give the cache lint the direction it lacked (#11518)
0084963 Daniel Han committed 2 months ago
Read the picker's orderRecommendedRows arguments as a set (#11663)
1a7a97a Daniel Han committed 2 months ago
Name forks after the chat they came from, and close their inherited history (#11607)
76846f4 Michael Han committed 2 months ago
Give every GGUF cancel-drain test the loaded-runner budget, and make the drain check deterministic (#11662)
e64fd8f Daniel Han committed 2 months ago
Recompute remote code's non-persistent buffers that transformers 5 leaves uninitialised (#11609)
c78f51c Daniel Han committed 2 months ago
Stall the signalled launchers in slices, so a signal at READY is not deferred (#11661)
ff29fb4 Daniel Han committed 2 months ago
Unsloth Studio: raise the huggingface_hub floor to what the pinned Diffusers main build needs (#11625)
78fbee7 Leo Borcherding committed 2 months ago
Export the trained weights when a full fine-tune loaded from a local folder is saved to GGUF (#11601)
ae3077a Nilay committed 2 months ago
Studio: read web pages in the encoding the page itself declares (#11598)
06c7663 Nilay committed 2 months ago
Studio: keep AI Assist label names and system prompt when the dataset format is Alpaca (#11597)
43e0921 Nilay committed 2 months ago
Studio: keep each imported tool call's own result when a chat file reuses call ids (#11596)
dcbb66e Nilay committed 2 months ago
Bound the unreadable-sample stall test by the fault it guards, not by halfway (#11655)
316e191 Daniel Han committed 2 months ago
Studio: keep the segmented control track visible in the dark Settings pane (#11653)
ffa4ff6 Michael Han committed 2 months ago
Make Phi-4-reasoning-vision load and preprocess images on transformers 5 (#11469)
310fb74 Daniel Han committed 2 months ago
Studio: sort the image picker by the chosen Hub sort, keep unsloth rows on top, and drop the Recommended sort (#11642)
fd1adc8 Michael Han committed 2 months ago
Studio: keep the model selector on the left, and centre its name on the icon (#11656)
7f3a147 Michael Han committed 2 months ago
Studio: copy a usable API key prefix, and show a new key from its start (#11652)
49752f6 Michael Han committed 2 months ago
Wait for the account-switch reload without evaluating across it (#11645)
89a3bf8 Daniel Han committed 2 months ago
Studio: steady the Images and Video loading spinners, and tidy the progress card (#11643)
62cc1e0 Michael Han committed 2 months ago
Close the WAL keepers a backend test opened before the next one runs (#11633)
3fec461 Daniel Han committed 2 months ago
Record the settings search flash instead of polling for it (#11632)
1c2612c Daniel Han committed 2 months ago