SageMaker Python SDK takes much time to submit the job on MacOS
Describe the bug When running sagemaker python sdk (latest version) on macos, it takes around 25-30 mins to submit the job. No logging or debugging is shown to help tracking the source of this slow performance. Any ideas? Thank you
Hi @achamma723, would you be able to share more info? Are you using a specific Framework or Algo Estimator or ModelTrainer? Are you the PySDK in a local jupyter server or IDE?
Hi @achamma723, would you be able to share more info? Are you using a specific Framework or Algo Estimator or ModelTrainer? Are you the PySDK in a local jupyter server or IDE?
Hello @benieric, I'm using the HuggingFace estimator from sagemaker latest version under the python 3.10 or 3.11 (tried both to check for the fix). I'm running it inside a local jupyter notebook in order to submit a training job. Its taking forever (tried a while ago, took more than 1 hour without any job creation)
Seems interesting error, can you share more info on your setup? Like what sagemaker and jupyterlab version?
I am on mac, firefox browser, and seems to work for me:
> pip freeze | grep jupyterlab
jupyterlab==4.3.6
jupyterlab_pygments==0.3.0
jupyterlab_server==2.27.3
jupyterlab_widgets==3.0.11
> pip freeze | grep sagemaker=
sagemaker==2.243.0
Hello @benieric, thank you for following this issue. As for the configuration, jupyterlab==4.3.3 jupyterlab_pygments==0.3.0 jupyterlab_server==2.27.3 jupyterlab_widgets==3.0.13 sagemaker==2.235.2 sagemaker-core==1.0.16
Also in my case, I'm trying to run a training job with a local model (and the ethernet connection is of high-speed). Yet still taking a good amount of time (sometimes 1h) before showing anything