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 modules_to_save trainable when reloading a LoRA checkpoint (#13303)
7d99558 Daniel Han committed 1 month ago
Let a torchcodec that fails to load its native library fall back instead of breaking import (#13348)
b43df24 Daniel Han committed 1 month ago
Mask packed document boundary labels for TRL older than 0.23.1 (#13324)
7aa0761 Daniel Han committed 1 month ago
Studio: stop a chat switch mid-reply crashing the app on a stale message index (#13352)
5fb2e85 Daniel Han committed 1 month ago
Keep token fields when DiffusionTrainer converts a config (#13335)
c16236b Vineeth Sai Varikuntla committed 1 month ago
Studio: continue LoRA training from a saved adapter checkpoint (#13158)
a5c7d49 Indrajit swain committed 1 month ago
Studio: stop offscreen reasoning transcripts listening to thread scroll (#13046)
73e1b5b Daniel Han committed 1 month ago
Studio: load and run DeepSeek-OCR and DeepSeek-OCR-2 (#13310)
dabf5ab Daniel Han committed 1 month ago
Studio: keep prompt cache on seeded tool-loop continuations (#10707)
243af4a Indrajit swain committed 1 month ago
Save LoRA-GA adapters relative to the original base so they reload correctly (#13322)
14b634b Daniel Han committed 1 month ago
AsyncGRPOTrainer with fast_inference: train and generate on one GPU with shared weights (#13292)
e1d6354 Daniel Han committed 1 month ago
Studio: Recommended Hub filter and chat picker list FP8 / NVFP4 on capable GPUs, GGUF lists every GGUF (#13240)
15c9d53 Michael Han committed 1 month ago
Studio: classify qwen_image21 GGUFs as text-to-image (#13333)
a4b9c62 Abhishek Acharya committed 1 month ago
Studio: use a ComfyUI text_encoders file for a local ComfyUI DiT instead of downloading the base repo's encoder (#13314)
a5403a2 Daniel Han committed 1 month ago
Studio: make Search and Code toggle states clear in dark palettes (#12108)
7bd0ebf UmranPros committed 1 month ago
Studio: let YuE2 finish a song instead of cutting it at the asked length (#13346)
85ac36d Daniel Han committed 1 month ago
Studio: fit an oversized DiffusionGemma GGUF to the GPU in Auto mode (#13312)
fcee648 Daniel Han committed 1 month ago
Studio: launch q4_0 instead of iq4_nl KV cache on CUDA/HIP GPUs (#13330)
caa9fbd Daniel Han committed 1 month ago
Studio: skip the SNAC detokenize probe when the ids are past the vocab (#13297)
0abda7a Daniel Han committed 1 month ago
Studio: install the Vulkan audio.cpp build on AMD GPUs (#13298)
0a6812e Daniel Han committed 1 month ago
Studio: let a text Expression column in Data Recipes render empty (#13296)
e7cb127 Daniel Han committed 1 month ago
Keep offloaded embeddings in host RAM when Trainer and accelerate place the model (#13320)
7c4061e Daniel Han committed 1 month ago
Studio: explain GGUFs the bundled llama.cpp cannot read, and default Ternary Bonsai to its loadable pack (#13304)
83a761b Daniel Han committed 1 month ago
Studio (AMD/ROCm): use MIOpen immediate mode for image and video convs (#13323)
ef5e050 Daniel Han committed 1 month ago
Studio: stop listing Bonsai DSpark drafter files as models (#11265)
902ef70 kevin committed 1 month ago