deepspeedai / deepspeedai/DeepSpeed

[REQUEST] Add automatic logging of parallelism and ZeRO config to WandbMonitor

Open
#7,494 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement
Dominant language
Python
Stars
43.1k
Forks
5k
Avg merge
4d 15h
Merged PRs (30d)
112

Description

Description

Currently, WandbMonitor in deepspeed/monitor/wandb.py initializes W&B and supports logging.
I would like to suggest adding a simple method (or automatic hook) in WandbMonitor that updates the W&B config with the core distributed training settings after initialization is complete. (Parallelism Rank & Zero Config)

Proposed approach

Add a method such as monitor.update_config(engine) (or update_config_once()) that extracts these values from the DeepSpeed engine (or other getter method) and updates W&B config once the engine is fully initialized.

This should keep overhead minimal while adding significant value (experiment settings) for users using distributed learning.

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 in deepspeed/monitor/wandb.py and trace WandbMonitor initialization alongside the DeepSpeed engine initialization mentioned in the request. Identify which parallelism ranks and ZeRO configuration values are available after initialization. Done means W&B config is updated once with those distributed-training settings while keeping overhead minimal.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
distributed-systems, machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
42/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.