aws-neuron / aws-neuron-sagemaker-samples Goto Github PK

The random_sentence() function used for benchmarking in the "inference/inf2-bert-on-sagemaker/inf2_bert_sagemaker.ipynb" notebook contains some potentially offensive random terms. Can we please replace those?

The amazon_reviews_multi is no longer available on HuggingFace (https://huggingface.co/datasets/amazon_reviews_multi/discussions/4#64c3898db63057f1fd3ce1a0). So, we'll need to use a different dataset for this training example:

training/trn1-bert-fine-tuning-on-sagemaker/bert-base-uncased-amazon-reviews.ipynb

Hi, I'm following the sample here to try to compile a model to Neuron and deploy on SageMaker.

Following the steps in the sample exactly, I am able to deploy the model, but it when I try to use it I get the 500 error and my CloudWatch traceback shows the following:

I only saw this error when trying to use a second model in the same instance while another one is running, but that should not be the case here.

To get this notebook to work, I had to update the ECR SDK version to 2.18.1 and I had to update the IAM role with a policy that allows SageMaker to get metrics for the benchmarking section. Would it be possible to update the notebook to include those details / instructions?