This model work by feeding multi sampling frame from video or raw video file?
#7
by CT-Ati - opened
If It is the multi video base
Can we have preprocessing script or optimum sampling rate If we want to apply this into streamming input src?
BTW I just assume base on this info in readme
Video preprocessing
The custom modeling code sets these env vars internally (matches the training-time setup). If you want to override them, set them in your shell before importing transformers:
Env var Default What it does
FORCE_QWENVL_VIDEO_READER torchcodec Video decoder backend
VIDEO_MAX_PIXELS 200704 Max pixels per frame (~448×448)
FPS 2.0 Frame sampling rate
FPS_MAX_FRAMES 240 Cap on total frames (covers ~2 min videos)
FPS_MIN_FRAMES 4 Floor for very short videos
env var that are specified by default are the optimum value, the model was predominately trained at 2 fps.