NVIDIA / NVIDIA/cuEmbed

About a big embedding table

Open
#9 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Cuda
Stars
50
Forks
9
PR merge metrics
No merged PRs in 30d

Description

We run cuembed_test.py with k=5000000(embedding nums) and n=7000(indices nums), and the results are:


cuEmbed: 0.0005564689636230469
Pytorch: 0.00020170211791992188
fprop test pass = tensor(True, device='cuda:0')
bprop test pass = False


it seems that cuEmbed is much slower than pytorch.

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 running cuembed_test.py with k=5000000 and n=7000, then compare the cuEmbed and PyTorch timings. Investigate the reported bprop test failure alongside the fprop result; done means the backward test passes and the performance difference is explained or addressed.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.