Add free-tier user limitation note: Fast mode can generate ~6s two-person video max d0973b6 C4G-HKUST commited on 26 days ago
Update generation mode descriptions: clarify GPU budget is up to limit, add 10s video generation note for 720s+40 steps 864d14b C4G-HKUST commited on 26 days ago
Add two generation modes: Fast (240s, 15 steps) and Quality (720s, custom steps) 37e0f4b C4G-HKUST commited on 26 days ago
Ensure generation_time_display always has valid value to prevent JSON decode errors b197409 C4G-HKUST commited on 26 days ago
Disable cache for gr.Examples to fix JSON decode error on repeated example clicks 62aa7d4 C4G-HKUST commited on 26 days ago
Update GPU duration to 720s and 2p-2 generation time to ~11 minutes a3f3eee C4G-HKUST commited on 26 days ago
Add note about 40 denoising steps in generation time display 5440852 C4G-HKUST commited on 26 days ago
Fix JSON decode error: add generation_time_display to outputs to match inputs for cache compatibility 3049f55 C4G-HKUST commited on 26 days ago
Fix examples: move generation time to second column and fix list index out of range error e9633ab C4G-HKUST commited on 26 days ago
Fix wav2vec model path to use Hugging Face model ID: facebook/wav2vec2-base-960h b2737a3 C4G-HKUST commited on 26 days ago
Fix VAE scale tensor device mismatch: recreate scale list on GPU and move T5 encoder c2db27b C4G-HKUST commited on 26 days ago
Move models from CPU to GPU in worker process (like LivePortrait) f2436d3 C4G-HKUST commited on 26 days ago
Simplify GPU handling following LivePortrait pattern: CPU in main process, GPU in worker via @spaces.GPU 7b63b72 C4G-HKUST commited on 26 days ago
Fix Stateless GPU environment: avoid CUDA init in main process, initialize GPU in worker process f0617b0 C4G-HKUST commited on 26 days ago
Track GPU device during initialization and use same GPU in worker process and inference 3c322ce C4G-HKUST commited on 26 days ago
Add GPU availability check in worker process for @spaces.GPU decorator 694fec8 C4G-HKUST commited on 26 days ago
Fix Gradio version and launch config: downgrade to 5.0.x, remove show_api parameter 5cb609b C4G-HKUST commited on 26 days ago
Update: increase GPU duration to 360s, add gradio==5.42.0, update README 7569450 C4G-HKUST commited on 26 days ago
Use @spaces.GPU decorator for GPU management (reference LivePortrait) f1ae228 C4G-HKUST commited on 26 days ago
Fix GPU detection and CPU fallback: remove ssr_mode, add runtime GPU check, support CPU mode for FaceInference b7867cd C4G-HKUST commited on 26 days ago
Fix Gradio launch config for HF Spaces (remove no_proxy, use queue) ef30698 C4G-HKUST commited on 26 days ago
Use flash_attn 2.7.4.post1 precompiled wheel (torch 2.6 compatible) 28f2de3 zzz66 commited on 27 days ago
