CSVLogger

torchrl.record.loggers.csv.CSVLogger(exp_name: str, log_dir: str | None = None, video_format: str = 'pt', video_fps: int = 30) → None

A minimal-dependency CSV logger.

Parameters:

• exp_name (str) – The name of the experiment.

• log_dir (str or Path, optional) – where the experiment should be saved. Defaults to <cur_dir>/csv_logs.

• video_format (str, optional) – how videos should be saved when calling add_video(). Must be one of "pt" (video saved as a video_<tag>_<step>.pt file with torch.save), "memmap" (video saved as a video_<tag>_<step>.memmap file with MemoryMappedTensor), "mp4" (video saved as a video_<tag>_<step>.mp4 file, requires torchvision to be installed). Defaults to "pt".

• video_fps (int, optional) – the video frames-per-seconds if video_format=”mp4”. Defaults to 30.