zjr2000 / GVL

Official implementation for paper Learning Grounded Vision-Language Representation for Versatile Understanding in Untrimmed Videos
https://arxiv.org/abs/2303.06378
MIT License
26 stars 6 forks source link

How to do Reinforcement Learning for DVC #6

Open DoigtByou opened 11 months ago

DoigtByou commented 11 months ago

Reinforcement learning is described in the paper,but I couldn't find it in the code.Can you help me find the reinforcement learning in the code? Thank you~

zjr2000 commented 11 months ago

Sorry for the late response, we have provided a config for reinforcement learning, you refer to this cfg to perform reinforcement training

DoigtByou commented 11 months ago

Thank you for your answer, but there are still some things I don't understand.For example,This is shown in the screenshot below.If the caption_decoder_type is 'standard',then the ‘flag’ will be false,whether the following 'if flag' statement is not taken, and whether the reinforcement learning code in the following if statement will not be executed. ![Uploading 1701844144(1).png…]()

zjr2000 commented 11 months ago

Thank you for pointing that out. I revisited my code and realized that I overlooked some elements related to rl during my previous cleanup. I have now updated the code accordingly. Your feedback was invaluable in identifying this oversight. Please see https://github.com/zjr2000/GVL/commit/91c237ed80168eac4675248abb47dff7fdbee7dc for details