Error for training job failed. reason: algorithmerror: exit code: 127

Hello,

Same to #969

I was training a DistilBERT model on SageMaker instance using fast-bert. I am using the ml.p2.xlarge instance for GPU processing.

When the function downloads the training image from ECR during **fit()**, I happen to receive _**"/usr/bin/env: ‘python\r’: No such file or directory".**_ See below -


And, at the end of stack-trace received the following - _**error for training job failed. reason: algorithmerror: exit code: 127**_


**Tech Stack-**

[fast-bert](https://github.com/kaushaltrivedi/fast-bert/tree/master/fast_bert) docker image
SageMaker NB Instance - ml.t2.medium
GPU Compute - ml.p2.xlarge

What could be the reason for this error? My IAM role has all the required permissions.

Kindly help.

Start by comparing this report with #969 and tracing the fast-bert Docker image invocation during SageMaker fit(). Confirm whether the reported `/usr/bin/env` failure is reproducible on the stated instances, then document the verified cause and resolution.

Written by the indexing model from the issue text.