Wrong results in Action Recognition task.

Hi!
I have tried Internvideo2-1B-clip in the action recognition task on K400 dataset, I try to use the model without the dataset class you designed.
So what I do in vision is catching 8 frames from video, transform it using test_transform, feed the processed clip into the vision encoder to get the 1x768 feature.
In text I just use the k400_categories.txt and kinetics_prompt you offered, after the text encoder it's 400x16x768 features.
Finally I get these two features in get_sim, and get a rank of the categories, but the result is very bad.
the answer is always not in the top5 choices, the model seems to randomly rank the categories.
I don't know if there is any wrong.
the model I use is chinese_alpaca_lora_7b, InternVideo2-stage2_1b-224p-f4.pt, internvl_c_13b_224px.pth, InternVideo2_CLIP_1B.pth.

Start by comparing the reported manual pipeline with the project's designed dataset class, focusing on test_transform, the vision and text encoders, get_sim, k400_categories.txt, and kinetics_prompt. Reproduce the K400 ranking with the listed checkpoints and verify the input shapes and checkpoint compatibility. Done means identifying the discrepancy and documenting a reproducible corrected result.