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: load a 16-bit Continued Pretraining run in 16-bit in Chat (#11480)
aa745b8 Nilay committed 2 months ago
Keep the compiler on for native architectures loaded with trust_remote_code = True (#11516)
7c28d74 Daniel Han committed 2 months ago
Studio: replace reasoning pagination with a continuous transcript (#11522)
5c63c8f Wasim Yousef Said committed 2 months ago
Install a prebuilt wheel's declared runtime dependencies before smoke importing it (#11536)
ba40410 Daniel Han committed 2 months ago
Warn when the installed transformers drops a pre-quantized checkpoint's bnb quant state (#11450)
b7aca55 Daniel Han committed 2 months ago
fix(grpo): autocast with DEVICE_TYPE_TORCH instead of a pinned "cuda" (#11461)
5f2136a li-lizhe committed 2 months ago
Studio: stop one audio chat from blocking every other chat (#11483)
aabbadb Nilay committed 2 months ago
Resolve the auto class for omni checkpoints, and guard embedding lookups (#11467)
3945bde Daniel Han committed 2 months ago
Add the Qwen-Image-2.1 family, and the Diffusers it needs by default (#11405)
c121e80 Daniel Han committed 2 months ago
Studio: warn when the disk is running low, and point at the caches (#11439)
18bc510 Daniel Han committed 2 months ago
Let a hosted pre-cast text encoder be safetensors, not only a pickle (#11454)
0b6bfed Daniel Han committed 2 months ago
Rework Studio dark mode onto a single surface and text scale (#11502)
83c5278 Michael Han committed 2 months ago
Let a pre-quant checkpoint be safetensors instead of a pickle (#11394)
0275f7e Daniel Han committed 2 months ago
Docker: pass JUPYTER_PORT and the notebook skip options through run.sh (#11484)
38ffe9a Nilay committed 2 months ago
Studio: chat with a full fine-tune in 16-bit instead of a 4-bit copy (#11479)
3ecdfa3 Nilay committed 2 months ago
Studio tool loop: close two gaps in the static screens that gate model-written code (#11399)
de3bb5e Daniel Han committed 2 months ago
Studio: validate the reasoning controls instead of coercing them (#9595)
3ace662 Apoze committed 2 months ago
Report a Linux ROCm APU host-backed pool of zero as measured, not unknown (#11451)
cf0e577 Daniel Han committed 2 months ago
Find a GGUF companion published after the weights on the chat load route (#11448)
004351b Daniel Han committed 2 months ago
Do not auto-enable padding-free for a model whose forward cannot take it (#11468)
2390f4f Daniel Han committed 2 months ago
Install the import-time Python dependencies before smoke testing a prebuilt wheel (#11527)
66f8495 Daniel Han committed 2 months ago
Keep generated RL configs on the reentrant gradient checkpoint path (#11519)
9f3153c Daniel Han committed 2 months ago
Skip TRL's redundant kbit prep for models Unsloth already prepared (#11494)
c960baa Daniel Han committed 2 months ago
Give the Docker step harnesses the environment those steps now declare (#11521)
2656ffe Daniel Han committed 2 months ago
Let the cache row's path label follow the interface font size (#11520)
ed15309 Daniel Han committed 2 months ago
Build and publish signed prebuilt CUDA 13 wheels for torch 2.13 and 2.14 (#11515)
a868d22 Daniel Han committed 2 months ago
Studio: resolve upstream llama.cpp releases, assets and digests without the GitHub API (#5886)
1fd6b42 Daniel Han committed 2 months ago