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: custom video sizes and Images-style prompt fields on Create videos (#14050)
2224182 Michael Han committed 1 month ago
Studio: list every GGUF build that shares a quant in a custom folder (#13957)
db0e78c oobabooga committed 1 month ago
AMD: Strix defers only to an installable generic tag; gfx1153 health gate classified at inventory refresh (#14022)
983226f Daniel Han committed 1 month ago
Token budget: weight split entropy by labels on TRL 1.3 to 1.14 (#14024)
b562d5d Daniel Han committed 1 month ago
Follow-ups to the GRPO reward locals and FSDP2 mixed precision merges (#14023)
029bc4a Daniel Han committed 1 month ago
Context parallel: gather the FSDP2 state dict in save_model on transformers 4.x (#14028)
88ac9b0 Daniel Han committed 1 month ago
FP8 GEMV on Blackwell + Triton 3.3: no graph break under torch 2.7, keep the exact GEMV shapes that are faster (#14026)
15eaf67 Daniel Han committed 1 month ago
Move the Unsloth Studio MLX stack to mlx 0.32.4 and mlx-lm 0.32.2 (#13992)
5021249 Long Yixing committed 1 month ago
perf(studio): enter Zoo's inference scope registry on Unsloth Studio's MLX paths (#13975)
a50be79 Long Yixing committed 1 month ago
Optimizer state offload: keep inactive state on the host for unchunked Q-GaLore weight_quant steps (#14025)
15e4bc8 Daniel Han committed 1 month ago
Windows hipinfo: an all-hiding ROCR mask still means no ROCm target (#14019)
6a33aad Daniel Han committed 1 month ago
Kernel follow-ups: norm launch on HIP and under compile, grouped GEMM frozen operands, no CE eviction hint on H100 (#13967)
ae99ca4 Daniel Han committed 1 month ago
GRPO: install MoE routing replay when UNSLOTH_GRPO_ROUTING_REPLAY=1 (#13916)
9f348f6 Daniel Han committed 1 month ago
GRPO: score old and reference logprobs over the truncated sampling support when opted in (#13900)
65d1fdb Daniel Han committed 1 month ago
Studio: faster frontend type check with TypeScript 7 and bun check (#13986)
03fcdea Etherl committed 1 month ago
Studio: improve AMD NPU discovery, CLI support, and chat feedback (#13797)
2de9425 oobabooga committed 1 month ago
Keep opted-in bnb 4-bit MoE experts outside the compiled block (#13946)
cb7a047 Daniel Han committed 1 month ago
RMSNorm: restore in-place dX for pre-norm decoder norms (#13984)
133489e Daniel Han committed 1 month ago
Fused LoRA kernels: support lora_dropout and lora_bias (#13898)
df63cab Daniel Han committed 1 month ago
studiobench CI: boot Studio with update checks off so a mid-run llama.cpp release cannot cover the scene (#14007)
cd2b12f Daniel Han committed 1 month ago
PEFT MoE conversion snapshot: add beam, which transformers main now fuses as qwen2_moe (#14005)
955e2eb Daniel Han committed 1 month ago
Studio: raise the startup bundle budget by #13998's growth (#14004)
dbd6b3c Daniel Han committed 1 month ago
Add opt-in TFLOPs and MFU logging with LoRA aware FLOP counts (#13885)
a1d4daa Daniel Han committed 1 month ago
Token budget: weight split-chunk TRL metrics by token count (#13978)
f046663 Daniel Han committed 1 month ago
Studio: searching a model maker in the Model hub lists Unsloth's uploads of its models (#13998)
d4e005e Michael Han committed 1 month ago
Studio: model picker ranks Recommended by the Model hub's Latest ranking (#13993)
bcf7c81 Michael Han committed 1 month ago
Studio: a hidden Chat with files panel stays hidden after the first message (#13999)
4ebc263 Michael Han committed 1 month ago
Cover explicit RoPE position ids and compute the starting cache on CPU (#13976)
815829f Daniel Han committed 1 month ago
Studio: scale the expanded markdown table's text with the UI font size (#13996)
24e2438 Daniel Han committed 1 month ago