NVIDIA-NeMo / NVIDIA-NeMo/Gym

[checkpoint] Enable Indian Banking Gymnasium snapshots

Open
#3,239 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

area:environment feature
Dominant language
Python
Stars
1.2k
Forks
349
Avg merge
1d 21h
Merged PRs (30d)
318

Description

Parent

Child of #3219 and #3024.

Problem and state owner

resources_servers/indian_banking/app.py owns the synthetic banking world/database, task, dialogue history, and turn counters in generic Gymnasium session state.

Proposed adapter

Explicitly opt in only after validating the complete world state is JSON-safe and engine globals are reconstructible.

Implement explicit checkpoint opt-in, route classification, export/restore/retire hooks, execution-to-session binding, revision reporting, atomic replacement under attempt N+1, and stale-attempt fencing.

Acceptance criteria

  • Tool mutations and dialogue turns continue without cross-session leakage; cached judge/simulator state is not silently shared.
  • Checkpoint after multiple mutations, replace the server process, restore, and compare the next operation with an uninterrupted execution.
  • Corrupt or incomplete bulk state activates no sessions.
  • Attempt N cannot mutate or retire restored attempt N+1.
  • Terminal cleanup is idempotent and releases restored state exactly once.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with resources_servers/indian_banking/app.py and trace how synthetic world state, dialogue history, and turn counters enter generic Gymnasium session state. Then map the proposed checkpoint hooks and attempt lifecycle against the acceptance criteria. Done means safe export and restore across process replacement, no cross-session leakage, corrupt state activates nothing, stale attempts are fenced, and cleanup is idempotent.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
backend, infrastructure
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.