Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Weighted Samples #15

Open
lvoegtle opened this issue Sep 25, 2024 · 0 comments
Open

Weighted Samples #15

lvoegtle opened this issue Sep 25, 2024 · 0 comments
Labels

Comments

@lvoegtle
Copy link
Collaborator

Allow that samples themselves specify a weight to be used for sampling these.

A possible implementation could be:

  • Fill a buffer with buffer_size samples
  • From that buffer, pick one sample based on sampling weight, and yield that sample
  • Discard one random sample from the buffer based on inverse weight
  • Add two new samples to the buffer
    But this has the downside, that half of the samples is discarded although being loaded. Research for better algorithms.

The sample weight shall be configured in the task encoder.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

2 participants