I am training a RL Agent using the library: https://github.com/thu-ml/tianshou. This library allows visualizing results using Tensorboard. But, in cases where I have to resume training from a checkpoint, I cannot resume logging from the previous point and the logging starts again losing all the previous progress. I tried keeping the tfevent files but then the writer starts overwriting on the graph.
Is there any way that I could start logging into the tensorboard from the checkpoint?