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
Tokenizer loading: skip the unused slow tokenizer load when the fast one is returned as is (#15834)
862098e Daniel Han committed 23 days ago
Studio: keep the Markdown block split linear after an unmatched $$ (#15886)
a45e1ad Daniel Han committed 23 days ago
Studio setup.ps1: install bun only when the frontend is built (#15933)
6cfbb87 Daniel Han committed 23 days ago
Cleanup: remove 5 redundant pass statements that leave bytecode unchanged (#15790)
5acda0f Daniel Han committed 23 days ago
Studio: adopt the WAV save_from_media already transcoded instead of decoding it again (#15916)
e5acea1 Daniel Han committed 23 days ago
Studio: stop re-hashing kept spills and re-reading the record on every prune (#15884)
9e2da97 Daniel Han committed 23 days ago
Studio: treat unknown parent liveness as alive in the llama-server reaper (#15921)
ec7aba0 Daniel Han committed 23 days ago
Studio: wake admission waits on queue state changes instead of their next poll (#15885)
6165c44 Daniel Han committed 23 days ago
Studio: make Recipes lint validators run on Windows and under a gitignored temp root (#15929)
2a75436 Daniel Han committed 23 days ago
Studio: reuse recipe shard row counts while no shard changed (#15869)
97a2f18 Daniel Han committed 23 days ago
GRPO PrefixGrouper: build the shared-prefix layout from one host copy instead of per-token loops and per-segment device fills (#15835)
9d26414 Daniel Han committed 23 days ago
Studio: write each training metrics flush and its run progress in one transaction (#15871)
497c302 Daniel Han committed 23 days ago
Studio: score unclosed Markdown link destinations in linear time (#15811)
4eafe33 Daniel Han committed 23 days ago
Studio: find VLM rows without an image or caption from Arrow validity (#15787)
5ad31fe Daniel Han committed 23 days ago
Studio: write transcoded WAV blocks raw and fix the header once on close (#15917)
dfd90e6 Daniel Han committed 23 days ago
GRPO PrefixGrouper: import cleanly on torch 2.6 and keep it off below 2.7 with a log line (#15836)
edfcb3c Daniel Han committed 23 days ago
Studio: find a code block's backtick runs by regex when picking its fence (#15913)
c1c4cdc Daniel Han committed 23 days ago
Studio desktop: ask desktop-capabilities in the launch probe on a capability cache miss (#15823)
ec051ae Daniel Han committed 23 days ago
Studio: fetch ShareGPT+image row images concurrently, in row order (#15786)
f6f13c0 Daniel Han committed 23 days ago
Studio: reuse the VLM URL probe's downloads in the conversion (#15870)
6055083 Daniel Han committed 23 days ago
Studio: fetch up to four Hub files at once over HTTP when partials resume (#15781)
98b37c2 Daniel Han committed 23 days ago
Studio: copy Whisper timestamp tokens to the host once per window (#15860)
a859cf8 Daniel Han committed 23 days ago
Studio: index the training history's resumed-later check (#15872)
68bf374 Daniel Han committed 23 days ago
Studio: decode format-check preview rows once per request (#15867)
8abdb18 Daniel Han committed 23 days ago
Studio: Decision tab loads at once, System and settings polish, main-theme borders restored (#15919)
4d345ab Michael Han committed 23 days ago
Studio: one text and control width for every settings row (#15905)
7377ca2 Michael Han committed 23 days ago
Studio: add a Quickstart heading over the API example tabs (#15898)
fa7ad0d Michael Han committed 23 days ago
Studio tests: give managed engines ports another worker cannot be handed (#15915)
d035dff Daniel Han committed 23 days ago
Studio: stop the first Wan2.2-A14B render from waiting for the second expert's pin (#15906)
9258bfc Daniel Han committed 23 days ago
Studio diffusion: let a 12.7 GB host load FLUX.2-klein-9B by spilling the int8 denoiser to a scratch file (#15911)
2572ed5 Daniel Han committed 23 days ago