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
Studio: Library grid Sort menu, tidier toolbar menus and card icons (#11978)
5f6b249 Michael Han committed 1 month ago
Studio: compile the VAE decode for DiT families, from the NVFP4 time budget pass (#10889)
2f51b4c Daniel Han committed 1 month ago
Studio: steady pinned model drop line and pill-shaped project drop highlight (#11986)
dd53ac0 Michael Han committed 1 month ago
Studio: install flashinfer on demand for NVFP4 without moving torch (#11730)
11282f1 Daniel Han committed 1 month ago
Fix PEFT base export targets silently exporting the base model (#11781)
6c647ca xiao committed 1 month ago
Studio: read text, Markdown and HTML uploads in the encoding they were written in (#11896)
2552172 Lukas committed 1 month ago
Studio: keep inline HTML text in its line when parsing an upload (#11818)
dcd0e1c Lukas committed 1 month ago
Studio: NVFP4 flashinfer backend kernel items (device guard, persistent barrier, bias path, cached dispatch) (#10731)
4a45ad4 Daniel Han committed 1 month ago
Studio: per-layer NVFP4 image policies, flashinfer FP4 backend and a gated auto row (#10730)
e15d7a3 Daniel Han committed 1 month ago
Studio: whole-model NVFP4 for the video families with hosted pre-quantized denoisers (#10729)
8ecfd30 Daniel Han committed 1 month ago
Studio: minimal OS sandbox for Python and Terminal tools on Linux and macOS (#11209)
588c7df oobabooga committed 1 month ago
Load 4.x-era configs that transformers 5 strict validation rejects (Llama 4 attn_temperature_tuning) (#11836)
ff8e658 Daniel Han committed 1 month ago
Divide by gradient accumulation for forwards that take **kwargs but return a mean loss (#11898)
ee495e4 Daniel Han committed 1 month ago
Studio: run Qwen-Image-2.1 int8/fp8 on 24 GB cards with the transformer resident and the text encoder streamed (#11883)
f2a9375 Daniel Han committed 1 month ago
Load NVIDIA ModelOpt FP8 checkpoints through the transformers fp8 quantizer (#11592)
614f2d5 Daniel Han committed 1 month ago
Pin the compiled mask wrapper only when it is compiled (#11957)
7d1b02b Daniel Han committed 1 month ago
fix(studio): recover from a dead Metal GPU queue instead of failing every later request (#11383)
8d15d04 Long Yixing committed 1 month ago
Keep the training state intact across a standalone evaluate() / predict() (#11860)
2e29cac Daniel Han committed 1 month ago
fix(studio): retain GGUF quants across cache-folder switches (#11938)
7af9386 Wasim Yousef Said committed 1 month ago
Hand a composition with no forward of its own to its thinker (Qwen3-Omni) (#11523)
3395708 Daniel Han committed 1 month ago
Load and fine-tune LongCat-Flash-Lite-Sparse on transformers' longcat_flash (#11620)
cccba44 Daniel Han committed 1 month ago
Format the two files #11526 left off the formatter's fixed point (#11955)
0ff222f Daniel Han committed 1 month ago
Studio: open Images and Video previews in the Library viewer (#11776)
c089e04 Michael Han committed 1 month ago
Studio: Library settings, storage and sortable list columns (#11771)
953a2f4 Michael Han committed 1 month ago
Resolve remote-code model classes, reach per-expert submodules with LoRA, run the root dispatch hook on the embedding's device (Nemotron-Labs-Teacher) (#11543)
98fc7fe Daniel Han committed 1 month ago
Studio: Library page for files, media and fine-tunes (#11770)
d751773 Michael Han committed 1 month ago
Studio: Qwen-Image-2.1 RoPE in real arithmetic inside the compiled blocks (#11888)
e628a57 Daniel Han committed 1 month ago
Studio: static step skip for image models that keeps CUDA graphs (#11737)
66c7a81 Daniel Han committed 1 month ago