Add sentencepiece for T5Tokenizer

Pin diffusers==0.30.1 to avoid Dinov2WithRegistersConfig (not in transformers 4.44.2)

Match model README: transformers==4.44.2, plain from_pretrained, THUDM/CogVideoX-5b

Load tokenizer and text_encoder with transformers to fix T5 Placeholder load error

chore: reduce conformed versions

Use Gradio 5.x (get_token/whoami); drop huggingface_hub pin and debug logs

Pin huggingface_hub==0.22.2 for HfFolder; add debug instrumentation

Pin huggingface_hub for Gradio HfFolder compatibility

Add CogVideoX-5B cinematic video generation with ZeroGPU

initial commit