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
Keep masked head_dim 256 SDPA training off cuDNN attention on SM100 (torch 2.14 NaN grads) (#12344)
3e070b4 Daniel Han committed 1 month ago
Studio: let GGUF vision models see dark text in transparent images (#12310)
f66bb10 Nilay committed 1 month ago
scan_packages: refuse VCS, URL and local-path specs before pip download (#12357)
01e9961 Daniel Han committed 1 month ago
Studio: load LTX-2 and LTX-2.3 on the pinned transformers 5.5 (#12299)
894fc7f Daniel Han committed 1 month ago
Fix MiniMax-H3 compiled render on torch 2.12, 2.13 and 2.14 (#12326)
3012a76 Daniel Han committed 1 month ago
Write the Ollama Modelfile from the trained chat template (#12311)
31f8b04 Nilay committed 1 month ago
Honour FP8Linear.block_size in the patched FP8 forward (32x32 block checkpoints) (#12317)
d167b89 Daniel Han committed 1 month ago
Move the Unsloth Studio MLX pins to mlx 0.32.3 and mlx-vlm 0.7.4 (#12334)
f07d552 Long Yixing committed 1 month ago
Studio: fix the desktop island corner and border seam (#12285)
1c4fcba Lee Jackson committed 1 month ago
Studio: fix Qwen-Image-2.1 recompiles and Ideogram 4 attention and device errors on torch 2.11 to 2.14 (#12350)
a019826 Daniel Han committed 1 month ago
Studio: turn train on completions back on when leaving CPT (#12308)
340e1d2 Nilay committed 1 month ago
Widen the pinned child GPU mask when extra args name a companion device (#11823)
2046e53 jishnav committed 1 month ago
Keep frozen BatchNorm running stats fixed during LoRA training (#12319)
48f6f63 Daniel Han committed 1 month ago
Studio: catch the tensor split abort behind a gdb backtrace (#12293)
d34cccc oobabooga committed 1 month ago
Studio: count tokens for templates that refuse an empty chat (#12336)
0dba817 Michael Han committed 1 month ago
Force non-reentrant gradient checkpointing for DeepSeek-V4.1 (#12318)
adbf205 Daniel Han committed 1 month ago
Studio: isolate the Windows Terminal on MXC's default tier (#12328)
b1c522a Etherl committed 1 month ago
Keep packed INT4 layers off the fused inference kernel in train mode (#12349)
9f2f65f Daniel Han committed 1 month ago
Studio: report CUDA graphs as on once the deferred speed profile arms them (#12301)
5efd2d8 Daniel Han committed 1 month ago
Fix the ShareGPT mapping for Llama 3.1, Qwen and Gemma chat templates (#12314)
3ff585e Nilay committed 1 month ago
fix(studio): record the bit widths of prequantized MLX models (#11940)
6f70f90 Tom Mathews committed 1 month ago
Studio: keep menus and popovers below the desktop titlebar (#12348)
f0af859 Michael Han committed 1 month ago
Studio: let the Mac window be moved during an app update (#12316)
1f4aa14 Nilay committed 1 month ago
Studio: load plain FP8 encoder safetensors without torchao (#12333)
fa595c6 oobabooga committed 1 month ago
feat(studio): batch KV-quantized and TurboQuant MLX loads with prompt snapshot reuse (#12343)
89641cc Long Yixing committed 1 month ago
Studio: fix the faint seam beside the chat composer (#12304)
3999e5f Lee Jackson committed 1 month ago
Studio desktop: survive AppKit exceptions during event dispatch on macOS (#12277)
97dc0c4 Michael Han committed 1 month ago
Studio: chunk large Qwen-Image-2.1 attention queries on ROCm (#12335)
80eb763 oobabooga committed 1 month ago
Studio frontend test: find the More flyout by its props in any order (#12356)
a61240a Daniel Han committed 1 month ago