Spaces:
Running on Zero
Running on Zero
Commit History
Yesterday
-
Add 7 Anima artist-style LoRAs to the catalog Running
Verified signaturecb8084e -
Add 7 Anima artist-style LoRAs to the catalog
Verified signature8b6514e -
Arcane LoRA: add its arcane artstyle trigger
Verified signature61625f9 -
Add 7 Anima artist-style LoRAs to the catalog
Verified signature74f1e59 -
Add 7 Anima artist-style LoRAs to the catalog
Verified signature95bbd15 -
Allow Anima LoRAs together with the reference image
Verified signature1531ef2 -
LoRA stack: up to 3 LoRAs per request for Krea, Anima and Nova
Verified signature73d536e -
Anima and Nova LoRA picker (request-scoped hooks, checked against the layers at boot)
Verified signaturee3da2c7 -
Request-scoped LoraStore: parse LoRAs on the CPU, attach hooks only while a request runs
Verified signature66fa473 -
Add 135 Krea 2 Asian Character LoRAs with on-demand loading
Verified signature686f2ab -
Remove SDXL; add per-request Krea LoRAs (distill 4-step, realism, NiceGirls)
Verified signature0306c14 -
Add Nova Anime AM v6.0 (2.9B) as a second Anima pipeline
Verified signature8960504 -
SDXL: compel prompt weighting with A1111 syntax conversion
Verified signature5c505a3 -
Move the other models back to the CPU before each call
Verified signaturebbd664c -
Add SDXL 0.9; keep models on CPU and move only the selected one to the GPU
Verified signatureb4e2e41 -
Revert SDXL for now (all models failing with NVML assertion)
Verified signaturefb583f3 -
Add SDXL 0.9 (fp16)
Verified signature26a94b5
2 days ago
-
Krea merge: only norm layers stay fp32
Verified signature5cc1831 -
Krea: use the Quiho uncensored Turbo merge
Verified signature75e093a -
Replace FLUX.3 with Krea-2-Turbo
Verified signature5e7713c -
Anima: optional reference image (anima-ref-adapter v2)
Verified signatured27c2c8 -
Remove NovelAI Diffusion Anime V2 (unused by the website)
Verified signature67f5a2a -
Remove write probe
Verified signature44d5873 -
probe
Verified signatureb04caea
3 days ago
-
FLUX.3: add base mode (real guidance) and a GPU-time guard
Verified signature3414165 -
FLUX.3: drop the prompt-upsampler step so prompts are used as written
Verified signaturec4831f6 -
Upgrade stack (diffusers 0.42.1, gradio 6) and add FLUX.3 Image Klein
Verified signature90c6008 -
Add torchvision (diffusers Cosmos transformer needs it)
Verified signature96a4290 -
Add Anima V1 alongside NovelAI V2
Verified signature8b085b5 -
Use Gradio 5 so transformers 4 and diffusers clip-skip work
Verified signaturec248975 -
Pin transformers<5 so diffusers clip-skip works
Verified signature3ce3e5e -
Switch to official NovelAI Diffusion Anime V2
Verified signaturee780412 -
Replace Anything v4.5 and Boogu/Senko with NoobAI-XL 1.1
Verified signature3cdc915 -
Restore normal GPU request cap (120s)
Verified signaturef076444 -
TEMP: cap GPU request at 52s for a low-quota test
Verified signature638ebdd -
TEMP: cap GPU request at 66s for a low-quota test
Verified signaturedc502e1 -
Rename LoRA keys to the pipeline form so the Senko V5 LoRA loads
Verified signature464a773 -
Add Boogu Base + Senko-san NAI V5 LoRA alongside Anything v4.5
Verified signature834f20f -
Convert to image models Space: Anything v4.5
Verified signature32a574f
25 days ago
-
Use the platform-managed ZeroGPU spaces runtime
Verified signature99699b6 -
Set up Qwen 3.8 27B BF16 chat on ZeroGPU
Verified signature15ebc20 -
initial commit
Verified signature8750c06