mosaicml / mosaicml/streaming

Streaming Dataset Simulator stops working beyond 10,000 batches

Open
#877 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug
Dominant language
Python
Stars
1.6k
Forks
206
PR merge metrics
No merged PRs in 30d

Description

Environment

  • OS: Mac OS (Sequoia 15.2)
  • Hardware (GPU, or instance type): Macbook (CPU Box)
  • Streaming Version: 0.11.0

To reproduce

Enter in training duration == 100,000 batches. When training duration is > 10,000 (we tried setting training duration 100,000), the streaming simulator hangs. The terminal stops printing out progress per batch. Additionally, the plot in the UI hangs as well.

Here is what the terminal looks like when the simulator hangs:

Image

Here's what the UI looks like when the simulator hangs:

Image

he simulator hangs:

Expected behavior

I expect the simulator to work when the training duration is set at 100,000 like it does when the number of batches is 10,000 or less. Here is what the terminal looks like when the simulator is working.
Image

Here is what the UI looks like when the simulator is working:

Image

Additional context

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 by reproducing the Streaming Dataset Simulator hang on Mac OS Sequoia 15.2 with Streaming 0.11.0, comparing 10,000 and 100,000 batches. Observe both terminal progress and the UI plot, then trace the simulator entry point responsible for batch progress and plotting. Done means 100,000 batches continue producing terminal progress and the UI plot without hanging.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
data, machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.