third_party/InternVideo/InternVideo2/single_modality/INSTALL.md

Installation

Requirements

We mainly follow UMT to prepare the enviroment.

pip install -r requirements.txt

We follow UMT to set --epochs 201 to avoid the potential interrupt in the last epoch.

We observed accidental interrupt in the last epoch when conducted the pre-training experiments on V100 GPUs (PyTorch 1.6.0). This interrupt is caused by the scheduler of learning rate. We naively set --epochs 801 to walk away from issue.

Note

To run InternVideo2 pretraining, you have to prepare the weights of the InternVL-6B visual encoder and VideoMAEv2-g, and set the your_model_path in internvl_clip_vision.py and videomae.py.