deepspeedai / deepspeedai/DeepSpeed

Details about BERT Pre-training

Open
#2,206 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

Hi , DeepSpeed team. I want to know more details about the example described in BERT Pre-training

It took 8 hr 41 min with 4 DGX-2. I wonder how many epochs of this training? which dataset do it use?
Suppose it runs 40 epoch, the same as the origin BERT paper. One epoch takes 781s. Each GPU processes 50 samples/s with 512 sequence length. 781 * 50 * 64 is 2,500,000 samples. So the dataset container 2.5m samples?

Thanks in advance.

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 the linked BERT Pre-training tutorial and its reported 8 hr 41 min result. Check the training configuration and documented metrics to determine the epoch count and dataset, then verify the arithmetic against the stated hardware and throughput. Done means adding the missing details to the tutorial or providing a definitive documented answer.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
documentation, machine-learning
Issue type
Documentation
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.