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
FSDP2 save: one writer for a tensor parallel x FSDP2 model (#16181)
c622888 Daniel Han committed 17 days ago
Studio diffusion prequant cache: do not count cgroup shmem as reclaimable (#16154)
e557a7d Daniel Han committed 17 days ago
Packing: do not serve compiled inference_mode metadata to a training step (#16183)
a32528b Daniel Han committed 17 days ago
Studio video: load cached weights offline while the Hub is unreachable (#16152)
f427201 Daniel Han committed 17 days ago
FSDP2: average tp gradients over dp when every trainable param is ignored (#16180)
b788000 Daniel Han committed 17 days ago
Studio: carry FP8 / NVFP4 compile cache tags into render and compile threads (#16158)
d90ef5b Daniel Han committed 17 days ago
Prevent Desktop owner lockout before adding accounts (#13961)
304e6ce Maheswar Kumar committed 17 days ago
Flex small-smem tiles: keep float32 calls on the default tiles (#16172)
18ce989 Daniel Han committed 17 days ago
Studio OpenAI API: refuse unsupported weights in selected JSONL chats (#16153)
4565773 Daniel Han committed 17 days ago
Decode graphs: preserve eager cache growth and reject stale replay (#16191)
788967b Daniel Han committed 17 days ago
Studio: chat with LFM3-Audio-2B (text and speech in, text out) (#16243)
d533017 Leo Borcherding committed 17 days ago
FP8: strip activation_scale sources when matching the 16-bit save conversion (#16156)
1c23549 Daniel Han committed 17 days ago
Studio: keep style keywords inside quoted state labels in the Mermaid guard (#16133)
8c36486 Daniel Han committed 17 days ago
Studio fine-tuning queue: do not start jobs cancelled during the batching probe (#16151)
77e3fa5 Daniel Han committed 17 days ago
Studio sd.cpp install: treat ELF symbol-version mismatches as loader refusals (#16150)
d27af99 Daniel Han committed 17 days ago
feat(studio): draft MLX tool, video, grammar, KV and LoRA replies (#16441)
b0a6f4b Long Yixing committed 17 days ago
Studio: LFM3-Omni audio input, export, LoRA defaults and Stop for custom-code generate (#16242)
2548528 Leo Borcherding committed 17 days ago
Saving: write the source tokenizer_class instead of TokenizersBackend (#16104)
073f6e0 Leo Borcherding committed 18 days ago
Loader: fetch each config probe once per from_pretrained call (#15757)
ac48952 Daniel Han committed 18 days ago
Studio: overlap credential reads and preserve startup recovery (#16389)
daaa2e2 Daniel Han committed 18 days ago
Studio diffusion: turn off coordinate descent where Max pins reduction configs (#16269)
681e05c Daniel Han committed 18 days ago
Studio: name FFmpeg when a video GGUF can't decode, and shrink clips with the ffmpeg llama-server uses (#16098)
f4c7a8f Leo Borcherding committed 18 days ago
Desktop: overlap owner liveness probes and skip redundant preflight HTTP
027efa3 Daniel Han committed 18 days ago
Cleanup: remove six private helpers and constants in core that nothing calls (#15695)
a81803f Daniel Han committed 18 days ago
Saving: drop the training truncation and padding from saved tokenizers (#16102)
25262ce Leo Borcherding committed 18 days ago
Studio: recognise a worker VRAM release that finished before the first settle poll (#16404)
dfad9c5 Daniel Han committed 18 days ago
Studio: wake the dispatcher on stop instead of waiting out its queue poll (#16350)
879a00f Daniel Han committed 18 days ago