We've been experiencing an uptick in OpenSSL failures on the downloads for the Swift for TensorFlow nightly toolchains during our CI runs. Detecting a failure and adding a delayed retry seems to significantly cut down on failed CI runs.
This propagates that fix to the Dockerfiles used for our nightly benchmarks, which should hopefully make our nightly benchmarks more robust.
We've been experiencing an uptick in OpenSSL failures on the downloads for the Swift for TensorFlow nightly toolchains during our CI runs. Detecting a failure and adding a delayed retry seems to significantly cut down on failed CI runs.
This propagates that fix to the Dockerfiles used for our nightly benchmarks, which should hopefully make our nightly benchmarks more robust.