OCR model downloads are sometimes really slow #22
Comments
robertknight
added a commit
that referenced
this issue
Jan 1, 2024
Fix slow model downloads by moving models to a different bucket which has S3 Transfer Acceleration enabled. I couldn't enable this for the original bucket because AWS doesn't support S3TA for buckets with a period in their name. With the original URL I was able to reproduce sporadic slow downloads when downloading with ureq, reqwest and curl. Fixes #22
robertknight
added a commit
to robertknight/ocrs
that referenced
this issue
Jan 2, 2024
Fix slow model downloads by moving models to a different bucket which has S3 Transfer Acceleration enabled. I couldn't enable this for the original bucket because AWS doesn't support S3TA for buckets with a period in their name. With the original URL I was able to reproduce sporadic slow downloads when downloading with ureq, reqwest and curl. Fixes robertknight/rten#22
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I noticed while testing the initial ocrs-cli crate release that the step in the initial run which downloads models was sometimes really slow.
Steps to reproduce:
~/.cache/ocrsocrs <image>Expected result: Model download is same speed as if using curl or a browser to download the file
Actual result: Model download was sometimes much slower
Note this issue is not about optimizations to the infrastructure (eg. using a CDN) or making the actual model files smaller (eg. via reduced precision or quantization). Those are all things that should be done, but out of scope here.
The text was updated successfully, but these errors were encountered: