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
feat(studio): add a setting to hide the macOS menu bar icon (#11937)
2b20770 Tom Mathews committed 1 month ago
Honor layers_to_transform and finetune_last_n_layers with regex target_modules (#14021)
71bf799 Vineeth Sai Varikuntla committed 1 month ago
Studio: raise the startup bundle budget for #14247 (#14258)
46e4f50 Daniel Han committed 1 month ago
Studio: Developer mode opt-in for experimental CPU video generation (#14247)
26e246d Daniel Han committed 1 month ago
FSDP2 QLoRA: give the all-gathered 4-bit weight back its Params4bit class and 3D shape (#14204)
39efb78 Daniel Han committed 1 month ago
Studio: build the OpenAPI schema off the event loop (#14239)
a770de9 Daniel Han committed 1 month ago
pre-commit: cut the kwarg formatter's CPU 6.5x and stop the ruff fork bomb (#14174) (#14249)
d122ab5 Daniel Han committed 1 month ago
Studio: route clicks and scrolls to the browser page on macOS 27 (#14216)
88afa84 Daniel Han committed 1 month ago
Studio: keep code colour while scrolling long chats (#14242)
8de2de2 Daniel Han committed 1 month ago
Studio: keep the Read aloud speech model loaded beside the chat model (#14191)
f969957 Etherl committed 1 month ago
Qwen3 decode: size the RoPE read by the explicit position, not only the KV length (#14156)
3697c7d Daniel Han committed 1 month ago
Studio: choose working folders when creating projects (#13873)
e468ee3 Lee Jackson committed 1 month ago
FP8 GEMV long-indexing test: seeded inputs and a last-bit tolerance (#14220)
5f3a0a2 Daniel Han committed 1 month ago
RoPE: build grown caches on the GPU again, 32768 rows at a time (#14206)
0cf4929 Daniel Han committed 1 month ago
doctor --gpus: read NVML P2P enum constants as ints (#14164)
355985f Daniel Han committed 1 month ago
Optimizer state offload: track DTensor state loaded on the host when resuming (#14155)
d7eef8e Daniel Han committed 1 month ago
Studio: stage prebuilt installs in dirs that inherit the parent ACL on Windows (#14212)
2a4b498 Daniel Han committed 1 month ago
Prebuilt wheels: drop -lineinfo, unreachable SASS and flash-attn PTX; add extension-free mamba-ssm (#14142)
a9819dd Daniel Han committed 1 month ago
Attention: GQA never forces SDPA math, cuDNN-over-varlen on sm100 only, native before xformers (#14158)
e6c95c2 Daniel Han committed 1 month ago
Studio: restore Qwen causal-conv1d on demand, narrow fatal mamba-ssm, provision SSM kernels in export (#13971)
7729c37 Daniel Han committed 1 month ago
Block compile: keep the GatedDeltaNet self-check out of a non-reentrant checkpoint (#14195)
cefb057 oobabooga committed 1 month ago
fix(studio): reject oversized Metal alongside loads before startup (#14209)
2e95c7c Lee Jackson committed 1 month ago
Pass layers_to_transform, use_rslora and modules_to_save to LoraConfig on the SentenceTransformer fast encoder path (#14150)
1db40e5 Vineeth Sai Varikuntla committed 1 month ago
head_dim 512 attention: use the kernel for masks made under inference mode (#14202)
9faa995 Daniel Han committed 1 month ago
Context parallelism: drop the SDPA wrappers each ring step leaves in torch's registry (#14193)
7ca1b3d Daniel Han committed 1 month ago
unsloth run: pass llama-server short flags like -ctv, -dev, -ctxcp through whole (#14224)
37989cb Daniel Han committed 1 month ago
Studio: keep reading durable chat output while its events are being saved (#14135)
ffd6c30 Daniel Han committed 1 month ago
Studio: export token usage from Profile stats as CSV (#14213)
cdcc9e7 Daniel Han committed 1 month ago