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
Stop the CUDA spoof from deciding whether a GPU test skips (#10864)
9effb89 Daniel Han committed 2 months ago
Studio: accept zero-based shard numbering in cached model weights (#10857)
b36d753 Johannes committed 2 months ago
Studio: keep a stopped prompt in later chat turns (#10625)
855bfee Chaitanya Laxman committed 2 months ago
Give Repo tests (CPU) a budget it fits in, and name the patcher when it falls back (#10862)
ed64d90 Daniel Han committed 2 months ago
Track vLLM tags through 0.29, and pin the bitsandbytes symbols (#10858)
525a199 Daniel Han committed 2 months ago
fix: do not default-on tool-call nudging on the external loop (#10685)
a514d83 Apurv Gupta committed 2 months ago
Use the short RoPE factor at exactly original_max_position_embeddings (#10681)
9b3b2b1 Vineeth Sai Varikuntla committed 2 months ago
Studio: per-account isolation for shared installs and the desktop app (#10588)
7434717 Daniel Han committed 2 months ago
Re-format the two lines that drifted, so pre-commit stops failing on main (#10715)
0eedeee Daniel Han committed 2 months ago
Ask the vllm-free import tests in a clean interpreter, not in sys.modules (#10855)
fe74520 Daniel Han committed 2 months ago
Stop a comment in unsloth_zoo from silently disabling the patched SFT trainer (#10854)
89a10db Daniel Han committed 2 months ago
Unbreak main, and fix the five causes reddening the PR backlog (#10832)
fc66948 Daniel Han committed 2 months ago
Docker Studio: stop disabling Unsloth's TRL patches on GPU hosts (#10825)
4e926db oobabooga committed 2 months ago
Reject conflicting precision flags in FastLanguageModel (#10756)
7c63bc8 Nilay committed 2 months ago
Stop capping unsloth chat at 512 on any backend and /v1/chat/completions at 2048 on transformers/MLX (#10665)
bf87e29 Nilay committed 2 months ago
Windows installer: fix installs when the temp path contains a space (#10765)
b3d1940 oobabooga committed 2 months ago
fix(dataprep): don't append EOS to full-size mid-stride chunks (#10734)
91f85a4 Caio Theodoro committed 2 months ago
Studio: Show an error when chat history cannot be restored (#10761)
8ee07d6 Nilay committed 2 months ago
Studio: narrower desktop window and a sidebar edge trigger (#10706)
22bbff6 Michael Han committed 2 months ago
Studio: keep long reasoning responsive with bounded pages (#10717)
27a4ff6 Wasim Yousef Said committed 2 months ago
Upload merged models to the requested branch or pull request (#10760)
6a5b21b Nilay committed 2 months ago
Unbreak main: raise the startup transfer budget to the margin it shipped with (#10700)
0d9952f Daniel Han committed 2 months ago
MLX CI: give the llama-server health wait the same 180s budget as the other health waits (#10674)
baabc00 Daniel Han committed 2 months ago
Studio: stop dropping Deep Research's thinking toggle on a llama.cpp connection (#10678)
98dd8a9 Daniel de Araujo committed 2 months ago
Studio: stop ignoring the Memory tab's gradient checkpointing setting (#10666)
e726fb1 Nilay committed 2 months ago
Keep the surrounding {{ }} when stripping bos_token from a chat template (#10682)
6c54ae2 Vineeth Sai Varikuntla committed 2 months ago
Let a Deep Research handoff survive a preamble beside the tool call (#10677)
51e6344 Daniel de Araujo committed 2 months ago
Refuse to format when ruff cannot run, and stop the fixed-point guard skipping itself away in CI (#10726)
d0dbe90 Daniel Han committed 2 months ago