Why do I get different results every time I run demo_video_text_retrieval.ipynb, even though my video and text inputs haven't changed?

[LORDPQK]

Why do I get different results every time I run demo_video_text_retrieval.ipynb, even though my video and text inputs haven't changed?

[ShayekhBinIslam]

Hey @LORDPQK , how is the config set internvideo2_stage2_config.py ? Results shouldn't be different.

Best,
Islam

[XiangguanMu]

I solve this problem by #107 , you need to set your path of InternVideo2-stage2_1b-224p-f4.pt in internvideo2_stage2_config.py in two places: model.vision_encoder.pretrained and pretrained_path