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
Gemma2: keep softcapping attention under int32 indexing and fall back to eager if compile fails (#12154)
9c74453 Daniel Han committed 1 month ago
Studio: price MLX loads in the memory panel and fit an unpinned context to available memory (#10287)
9828490 Long Yixing committed 1 month ago
Studio: stop a hung system node or npm from stalling setup (#12165)
c176530 oobabooga committed 1 month ago
Studio: make compile knobs reach the render thread on torch 2.12+ (#12075)
a735bcc Daniel Han committed 1 month ago
Studio: add a Scroll while generating setting (Auto-scroll or Manual) (#12110)
e7a5b49 Michael Han committed 1 month ago
Studio: show when a prompt was sent while hovering it (#12170)
869af7c Michael Han committed 1 month ago
Keep scan_packages.py from tripping Bitdefender's Python stealer signature (#12169)
99c3928 Daniel Han committed 1 month ago
Fix Gemma and Gemma2 embeddings scaled twice on transformers 5.4+ (#12116)
88d35ac Daniel Han committed 1 month ago
Gemma2: use flash_attn_with_kvcache for cached decoding (#12112)
379f14a Daniel Han committed 1 month ago
Fix Gemma2 padding masks during batched cached decoding (#12008)
f7c9c68 cherry77-cloud committed 1 month ago
Windows installers find an NVIDIA GPU on the PCI bus, and say which CUDA it can use (#11166)
268b771 Daniel Han committed 1 month ago
Add a Windows probe for code integrity blocks, and audit bundle signatures in CI (#10408)
5a28557 Daniel Han committed 1 month ago
Remove the reflection-emit apparatus and all four emitted types (#11193)
bd612c8 Daniel Han committed 1 month ago
Recover CUDA compute capabilities without emitting a P/Invoke type (#11173)
b7d2b33 Daniel Han committed 1 month ago
Refresh a rewritten shortcut's icon where the shell cannot define the type (#11116)
ecc691b Daniel Han committed 1 month ago
Ask Python for the process image table when the native helper is unavailable (#11115)
7e107f8 Daniel Han committed 1 month ago
Unsloth Studio Installer: ask Python for a path identity before giving up on an exact one (#11104)
72acc86 Daniel Han committed 1 month ago
Fix rowwise FP8 scale axes in fused LoRA backward (#11799)
7581d50 cherry77-cloud committed 1 month ago
Studio: TurboQuant KV cache option for MLX inference (#11170)
eadcab6 Long Yixing committed 1 month ago
Fall back from FBGEMM for rowwise FP8 on GPUs it has no kernel for (RTX PRO 6000 / 5090) (#12098)
5d6ecc4 Daniel Han committed 1 month ago
Speed up block-FP8 LoRA training: run FP8 linears eagerly, 8 warps for 128-row GEMM tiles (#12027)
7c8e50a Daniel Han committed 1 month ago
Settle the permission step's reloads on the pill instead of networkidle (#12153)
d96e922 Daniel Han committed 1 month ago
Studio: keep checkpoint compaction under --disable-tools (#12119)
9290d18 Daniel Han committed 1 month ago
Studio: move Managed accounts to the Accounts tab and drop Mark as unread from chat menus (#12120)
b9716be Michael Han committed 1 month ago
Wait for every diffusion run a test started before undoing its runs dir (#12149)
8625c8d Daniel Han committed 1 month ago
fix(studio): mark /api reads no-store so an idle desktop app stops rewriting its disk cache (#12148)
1dddc14 alkinun committed 1 month ago
Point dsh at Unsloth through a --patch overlay instead of settings.yaml (#12145)
2c03020 Daniel Han committed 1 month ago
Batched serving on the MLX path: several replies decoding at once (#10310)
1232012 Long Yixing committed 1 month ago
Gate the Mllama CUDA forward test on has_real_cuda (#12143)
839cdbf Daniel Han committed 1 month ago