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: size music waits for Metal and Vulkan, not only CPU (#13551)
cb53b9b Daniel Han committed 1 month ago
unsloth start opencode: detect OpenCode V2 installed as plain opencode (#13550)
d127a5c Daniel Han committed 1 month ago
Studio: read ZE_AFFINITY_MASK the way Level Zero does (#13528)
ce24361 Daniel Han committed 1 month ago
Studio: let a model loaded at a set context reload on Auto (#13555)
2890f2c Michael Han committed 1 month ago
Studio: rounded-rectangle hover on the header model selector (#13546)
6d10726 Michael Han committed 1 month ago
Studio: rename the composer's Code toggle to Tools (#13402)
8afc771 Michael Han committed 1 month ago
Studio: XPU native bf16 no longer depends on intel-ocloc (#13527)
f6ed46d Daniel Han committed 1 month ago
Studio: transcribe with downloaded MLX speech-to-text models on Apple Silicon (#13443)
cf6e3e3 Daniel Han committed 1 month ago
Fix the exec-literal, encoding and formatter gates #13520, #13534 and #13536 left red on main (#13543)
428ac98 Daniel Han committed 1 month ago
Studio (AMD iGPU): split an iGPU row's total into VRAM and shared memory (#13421)
8a0e453 Daniel Han committed 1 month ago
Save int4 QAT LoRA models as real int4 with save_pretrained_torchao (#13413)
bb4398f Daniel Han committed 1 month ago
Studio: run the Shutdown cleanup off the event loop (#13534)
3b41804 Daniel Han committed 1 month ago
Studio: per-family diffusion tables move into the family registries (#13536)
e227a53 Daniel Han committed 1 month ago
Re-measure the Studio startup budget after the Library and composer merges (#13533)
f68585e Daniel Han committed 1 month ago
Studio: reveal and delete for LM Studio, Ollama, custom-folder and other local models (#13507)
090365d Michael Han committed 1 month ago
Studio installer (AMD, Linux): no generic ROCm torch for RDNA 1 / Polaris when the arch is readable (#13419)
53199ea Daniel Han committed 1 month ago
Studio: train Z-Image LoRAs on attention and feed-forward (#13233)
070a0a5 Daniel Han committed 1 month ago
Turn off Granite 3.0-3.2 attention dropout on float16 loads (#13417)
b4266cb Daniel Han committed 1 month ago
test_packing_sdpa: take the loss path it reads labels from (#13434)
3d35441 Daniel Han committed 1 month ago
Keep the saved Hugging Face login when from_pretrained gets a token (#13486)
5124404 Nilay committed 1 month ago
Format the diffusion auto-optimizer and LTX-2.5 prequant files (#13526)
4c9f6c7 Daniel Han committed 1 month ago
Studio: train image LoRAs with aspect-ratio buckets (#13218)
1e30dc7 Daniel Han committed 1 month ago
Studio: scale the SDXL loss in fp16 and keep image LoRA training off devices it cannot use (#13278)
34d8315 Daniel Han committed 1 month ago
Studio: random flip opt-in, plus checkpoint retention, caption dropout, LoRA dropout and timestep shift on the image Train tab (#13210)
daf65ab Daniel Han committed 1 month ago
Studio: diffusion flow_shift is the effective shift, and min-SNR is recorded as unused for flow-matching runs (#13215)
92fcec8 Daniel Han committed 1 month ago
Studio: open a project source from the Sources tab (#13441)
d7acd1d Daniel Han committed 1 month ago
Studio: show pending llama.cpp, whisper.cpp and audio.cpp updates on one card (#13497)
f2321e9 oobabooga committed 1 month ago