OOM During Model Evaluation

Hello,
 I tried running inference for a 120 frames video, but get an OOM on an H100 GPU, did you try to test your model on long videos with such frame length? What was your results? and how can I prevent such an error?


<img width="742" height="422" alt="Image" src="https://github.com/user-attachments/assets/045e87f3-ffe5-494e-992d-8c7c1ae120c9" />