Shortcuts

FlatStorageCheckpointer

class torchrl.data.replay_buffers.FlatStorageCheckpointer(done_keys=None, reward_keys=None)[source]

Saves the storage in a compact form, saving space on the TED format.

This class explicitly assumes and does NOT check that:

  • done states (including terminated and truncated) at the root are always False;

  • observations in the “next” tensordict are shifted by one step in the future (this is not the case when a multi-step transform is used for instance) unless done is True in which case the observation in (“next”, key) at time t and the one in key at time t+1 should not match.

Docs

Access comprehensive developer documentation for PyTorch

View Docs

Tutorials

Get in-depth tutorials for beginners and advanced developers

View Tutorials

Resources

Find development resources and get your questions answered

View Resources