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: flag a modules.json folder of only dots and slashes that leaves the model (#14527)
b515632 Daniel Han committed 29 days ago
Studio diffusion: pool SDXL textual inversion prompts at EOS (#14846)
2b052f0 Daniel Han committed 29 days ago
Studio H3: let Auto residency see the card again on ROCm and NVIDIA without nvidia-smi (#14820)
b8513d2 Daniel Han committed 29 days ago
Studio diffusion: compile MiniMax-H3 with torchao float8 weights (#14621)
22a8b5d Daniel Han committed 29 days ago
Tests: name text encodings and gate CUDA skips on has_real_cuda (#14871)
c2efedc Daniel Han committed 29 days ago
Studio: Install Training button in the Train hint and the Export banner (#14861)
d368a7b Daniel Han committed 29 days ago
Studio diffusion: free the failed group offload setup before the OOM retry (#14821)
fe80287 Daniel Han committed 29 days ago
Format two files merged without the repo formatter in #14834 and #14819 (#14878)
7ca9ad8 Daniel Han committed 29 days ago
Studio diffusion: latch the CFG-batch verdict per load and across processes so a seed keeps one path (#14849)
f15d723 Daniel Han committed 29 days ago
Studio browser: keep a temporary chat's hidden pages out of history (#14855)
71d24f7 Daniel Han committed 29 days ago
Desktop: keep the environment when WEBKIT_DISABLE_COMPOSITING_MODE=0 leaves compositing on (#14837)
d3414a8 Daniel Han committed 29 days ago
Studio updater: install a matched unsloth + unsloth-zoo pair when the index lags (#14834)
3cf1617 Daniel Han committed 29 days ago
Studio: read the prequant source etag from the repo blob under the shared Hub blob store (#14819)
ae01302 Daniel Han committed 29 days ago
Studio Images: run the Hires fix latent encode where an offloaded SDXL VAE computes (#14851)
82c4efe Daniel Han committed 29 days ago
Studio diffusion: keep the CFG-parallel weight probe inside large inference weights (#14844)
89ae8fa Daniel Han committed 29 days ago
Studio MLX images: release MLX's buffer cache after each generation (#14795)
7c26a1c Daniel Han committed 29 days ago
GPT-OSS template: keep earlier answers on the final channel when the turn has thinking (#14700)
02d9ba1 Daniel Han committed 29 days ago
Studio: fetch VLM dataset row image URLs through the SSRF-guarded fetch (#14701)
866fed8 Daniel Han committed 29 days ago
Fix Gemma 4 E2B / E4B DPO CheckpointError after the ref adapter switch (#14705)
88041a6 Daniel Han committed 29 days ago
Studio: only trust llama-server's /health when our own child holds the port (#14706)
32081f4 Daniel Han committed 29 days ago
Studio: API keys can't make a Data Recipe read credentials from the server environment (#14699)
1eab41f Daniel Han committed 29 days ago
Studio: hold the image render slot while a cold FLUX.3 upsampler start comes up (#14830)
2f9e7d3 Daniel Han committed 29 days ago
Studio: budget HunyuanImage-2.1's automatic int8 text encoder at its hosted file (#14833)
ee55c3c Daniel Han committed 29 days ago
Studio diffusion: hold cudnn.benchmark off for FLUX.2 klein so a seed decodes the same image in every process (#14770)
5c6673a Daniel Han committed 29 days ago
Studio MLX images: list Q5_0 / Q5_1 in the unsupported GGUF refusal (#14839)
4faa9f4 Daniel Han committed 29 days ago
Studio: pick the vision projector precision per device, and never F16 for MiMo-VL (#14679)
3af6835 Daniel Han committed 29 days ago
Studio diffusion: onload torchao int8 / fp8 weights through the pinned chunks (#14832)
ea9db57 Daniel Han committed 29 days ago
Format test_video_stream_encode.py, merged without the repo formatter in #14822 (#14852)
a9da74c Daniel Han committed 29 days ago
Studio: restore a kept media model's price after a failed replacement (#14826)
3e35cbb Daniel Han committed 29 days ago