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: show diffusion training preparation progress (#15337)
bafc1e8 Daniel Han committed 26 days ago
Tests: follow TRL's renamed tiny Phi-3 and Qwen3.5 repos (#15355)
e3a276a Daniel Han committed 26 days ago
Tensor parallel from UNSLOTH_TP_SIZE for unmodified training scripts, with Trainer checkpoints and 8-bit AdamW (#15326)
97a1732 Daniel Han committed 26 days ago
fix(studio): distinguish media RAM offloading from insufficient memory (#15341)
e48962c Daniel Han committed 26 days ago
Tests: name the encoding in the gloo export guard, and treat a bare device index as CUDA in the YaRN load check (#15351)
ba161f5 Daniel Han committed 26 days ago
Import without Triton: define Unsloth's kernels against the zoo triton stub, visible to unsloth only (#15346)
7083076 Daniel Han committed 26 days ago
Studio: exclude NVIDIA NPU from GPU memory accounting (#15334)
5c2b1fb Daniel Han committed 26 days ago
Trim comments in the Studio frontend and Tauri shell (#12966)
9e44983 Daniel Han committed 26 days ago
Studio diffusion: pin FLUX.2-klein's reduction-config filter on every arch (same seed, same image across processes) (#15340)
b760119 Daniel Han committed 26 days ago
Trim comments in the installers, scripts and CI config (#12969)
3c812a3 Daniel Han committed 26 days ago
Intel XPU: fix 4bit dequantize and gemv on plain (not double-quantized) absmax (#15322)
d6cb541 Daniel Han committed 26 days ago
torchrun: plain Trainer DDP find_unused default, FSDP2-collective SentenceTransformer save (#15324)
e0a8f07 Daniel Han committed 26 days ago
Studio: warn when a custom Hub endpoint redirects downloads to huggingface.co (#15329)
241bba9 Daniel Han committed 26 days ago
Compiled blocks: packed rows run the flash varlen kernel inside the graph (#15325)
85c9d20 Daniel Han committed 26 days ago
Fix TRL 0.27-0.28 GRPOTrainer import on transformers >= 5.3 (No module named 'weave') (#15343)
8ed7abb Daniel Han committed 26 days ago
Studio: keep Images, Video and Hub search typing responsive (#15306)
9e1ddce Daniel Han committed 26 days ago
Native FSDP2: memory-scalable 16-bit loading (each rank reads only its shard), and the fp16 start scale on sub-meshes (#15225)
20bbee4 Daniel Han committed 26 days ago
Studio: open the tile's own menu on a right-click in the Images strip (#15276)
12e8e47 Debojyoti Banerjee committed 26 days ago
Native FSDP2: fix generate after QAT convert (mixed torch.Tensor and DTensor) (#15311)
babbe6c Daniel Han committed 26 days ago
Studio diffusion: seed the small-host int8 placement from its measured peak (FLUX.2-klein on a 12.7 GB T4: 30.4 to 10.6 s per image) (#15330)
2b116f3 Daniel Han committed 26 days ago
Studio: Cosmos3-Nano and Cosmos3-Edge video families (#14554)
f1a1805 sbabamohamad committed 26 days ago
Studio video: write the offloaded native int8 build to the prequant cache after the load (Wan2.2-A14B first clip 97 s sooner) (#15331)
b759a5b Daniel Han committed 26 days ago
Studio: sample Vulkan drafts on the CPU when the vocab misaligns the logits rows (#15258)
f1fbcf8 Daniel Han committed 26 days ago
Fix GKD normalization after response generation (#15222)
b39fc88 cherry77-cloud committed 26 days ago
Align System header controls and avoid the close button (#15295)
619763f Lee Jackson committed 26 days ago