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 four POSIX-only test modules from failing to collect on Windows (#11457)
1d8a608 Daniel Han committed 2 months ago
Studio: isolate the sandbox traversal test from the real home directory (#11471)
75c5e3e Daniel Han committed 2 months ago
Read the sidebar and reasoning contracts as claims, not as literals (#11404)
47e595c Daniel Han committed 2 months ago
Studio: stop suggesting bnb-4bit models MLX cannot load on Mac (#9722)
b37bddb Nilay committed 2 months ago
Apply the logit scale on the fused cross entropy path (#11277)
917af38 Vineeth Sai Varikuntla committed 2 months ago
Studio: keep an AMD dGPU and a Vulkan choice through install and update (#11442)
6770853 Daniel Han committed 2 months ago
Studio: let the owner allow managed accounts private provider base URLs (#11436)
9ceb263 Daniel Han committed 2 months ago
Studio: prefer the discrete AMD GPU when an iGPU enumerates first on Linux (#11437)
17e1490 Daniel Han committed 2 months ago
Docker ROCm Studio image: ship JupyterLab with the notebooks, as the CUDA image does (#11286)
b19b41b Leo Borcherding committed 2 months ago
Point the two fallback tests at a temporary root Windows accepts (#11455)
468d550 Daniel Han committed 2 months ago
Stop SFTTrainer downcasting configs that subclass SFTConfig (#11440)
d48ab69 Daniel Han committed 2 months ago
studio: preserve standalone gguf settings in compare mode (#11311)
dfb130d Maheswar Kumar committed 2 months ago
Studio: edit a project from the Projects page, and sort it from the Updated column (#11443)
2980768 Michael Han committed 2 months ago
Offer FlexAttention when head_dim puts every flash kernel out of reach (#11102)
c99204e Daniel Han committed 2 months ago
[perf] Qwen3.5-9B LoRA SFT with unsloth-cli.py on a B200: 0.84 s to 0.77 s per step at 1 GPU, the model-independent half of #10744 (#11238)
a801cae Hexu Zhao committed 2 months ago
Refuse a compiler cache path shlex cannot parse, not only one with a space (#11414)
edadaf2 Daniel Han committed 2 months ago
Studio: three-state Thinking and Tool calls visibility in Display (#11433)
49b4843 Michael Han committed 2 months ago
fix(sft): honor max_length below the model context limit (#11266)
289daaa cherry77-cloud committed 2 months ago
Studio: require a published digest on upstream llama.cpp prebuilts too (#11402)
0ec1e92 Daniel Han committed 2 months ago
Studio: read the saved credential and its presence in one query (#11420)
64e4d52 Daniel Han committed 2 months ago
Studio: drop the "When context fills" setting and always use the server's policy (#11444)
9845b2a Michael Han committed 2 months ago
Studio: the drag-and-drop switches go, the gesture has one behaviour (#11441)
432b4db Michael Han committed 2 months ago
A failed presence check in the indicator driver should say what the page looked like (#10435)
fc86fd0 Daniel Han committed 2 months ago
Fail Source lint on a duplicate TypeScript import binding (#8548)
83fb493 Daniel Han committed 2 months ago
remove gguf shard controls introduced by #6107 (#11315)
ef6ae0b Maheswar Kumar committed 2 months ago
fix(studio): widen the link reference probe to match the definition probe (#11375)
949ce69 Daniel Han committed 2 months ago
Studio auth and bootstrap gates: throttle desktop-login, narrow the tunnel CORS origin, identify Colab's proxy (#11398)
da77eb8 Daniel Han committed 2 months ago
Studio: scope the GGUF converter pin to the conversion (#11417)
f341e82 Daniel Han committed 2 months ago
Derive the chat UI wall-clock watchdog from the turn timeout (#9025)
2c3873d Daniel Han committed 2 months ago