thu-ml/tianshou

Allow finding the corresponding episode from a sample in reply buffer

Open

#881 opened on Jun 1, 2023

View on GitHub
 (1 comment) (1 reaction) (0 assignees)Python (7,121 stars) (1,072 forks)batch import
RNNenhancementgood first issue

Description

  • I have marked all applicable categories:
    • exception-raising bug
    • RL algorithm bug
    • documentation request (i.e. "X is missing from the documentation.")
    • new feature request
  • I have visited the source website
  • I have searched through the issue tracker for duplicates
  • I have mentioned version numbers, operating system and environment, where applicable: 
    import tianshou, gymnasium as gym, torch, numpy, sys
print(tianshou.__version__, gym.__version__, torch.__version__, numpy.__version__, sys.version, sys.platform)

I'm trying to save the PrioritizedVectorReplayBuffer by its eposide. But I am not sure if there is a key asscioated with the eposide. Are there any way to find the eposide of a sample?

PrioritizedVectorReplayBuffer( info: Batch( env_id: array([0, 0, 0, ..., 0, 0, 0], dtype=int32), players: Batch( env_id: array([0, 0, 0, ..., 0, 0, 0], dtype=int32), ), lives: array([5, 5, 5, ..., 0, 0, 0], dtype=int32), reward: array([0., 0., 0., ..., 0., 0., 0.], dtype=float32), terminated: array([0, 0, 0, ..., 0, 0, 0], dtype=int32), elapsed_step: array([1, 2, 3, ..., 0, 0, 0], dtype=int32), ), act: array([2, 2, 2, ..., 0, 0, 0]), done: array([False, False, False, ..., False, False, False]), policy: Batch(), rew: array([0., 0., 0., ..., 0., 0., 0.]), terminated: array([False, False, False, ..., False, False, False]), obs: array([[]]), truncated: array([False, False, False, ..., False, False, False]), )

Contributor guide

Tech stack
pythonpytorch
Domain
machine learningdata
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
needs maintainer response
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
mostly clear
Prerequisites
Familiarity with Tianshou's replay buffer architectureBasic Python
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
60
Research direction
Examine the PrioritizedVectorReplayBuffer class, likely in `tianshou/data/buffer/prioritized.py`. Look for how episodes are stored in the `elapsed step` field or similar. The goal is to add a method (e.g., `get episode id(sample indices)`) that returns the episode number for each sample. Check if there is an existing mapping from sample to episode, or if it needs to be reconstructed from `done` and `terminated` flags. Consider how episodes are tracked in the vectorized buffer (multiple environments). The issue has no maintainer response yet, so first clarify the expected API and any edge cases.