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

fix: increase timeout rate of lora metadata/model downloads #211

Merged
merged 5 commits into from
Mar 7, 2024

Conversation

tazlin
Copy link
Member

@tazlin tazlin commented Mar 7, 2024

The new multiprocess worker paradigm is very sensitive to timing in certain respects. This PR seeks to address the tendency for LoRa metadata or model downloads to get hung up at certain points in the model manager, which tends to lead to the worker detecting the working process as stale and leading to all sorts of issues.

In short:

  • Adjusts timeouts to be more in line with real-time generation expectations as is the case with the worker.
  • Gives up metadata collection sooner in certain probably unrecoverable circumstances, such as getting a 500 or HTML doc back.
  • Increases the number of download threads to improve the chances that things move along
    • During testing, this is has been increased to 5x the previous number which greatly increases the rate at which LoRa based tests run.
  • Increasing the polling frequency of threads (THREAD_WAIT_TIME)
  • Binds several other time.sleep(...) calls to the value of THREAD_WAIT_TIME.

@db0
Copy link
Member

db0 commented Mar 7, 2024

Looks OK, my comment is more than I don't think increasing the threads would do much, but it won't hurt.

@tazlin tazlin merged commit e0fb477 into main Mar 7, 2024
2 checks passed
@tazlin tazlin deleted the lora-metadata-timeout branch March 7, 2024 16:44
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants