fix(app): handle dtype alignment for style/subject reference pipelines

fix(app): align input tensor dtypes with model dtypes during inference

Update app.py

Update app.py

add use low vram option

Update app.py

do not cache examples

Update app.py

disable low_vram mode

Update app.py

Update app.py

fix typo

Update app.py

Update app.py

Update app.py

Update app.py

Disable gradio analytics to save some memory

Update app.py

Update app.py

Update app.py

Update app.py

return PIL images everywhere, do not convert to PNG to avoid artifacts

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py