Deepracer : Locally trained model not valid for import on DeepRacer console

Question

Hi,
After uploaded my training model (local training) to my AWS S3 bucket, I always have got an error message "Invalid model" when I tried to import my model to AWS Deepracer Console for evaluating.
I verified the required files for model imports uploaded in S3 bucket specify in 
 https://docs.aws.amazon.com/deepracer/latest/developerguide/import-export-models.html and It'all OK.
So, from explication provider in "Model not valid", I do not understand what it means of "We can't validate your model because it's been edited... ".

Thanks for your help.

Accepted Answer

This typically happens because someone trained locally but exceeded the documented limitations of the action space.

Answer

Thank you very much for reply.
Sure enough, there were 3 points that had a speed greater than 4. Now this is fixed.

However, I have another error which is apparent when I restart a new training:

Unable to find deepracer checkpoint json
Exit because model_checkpoint_name is None of type 
s3 failed, retry after 1.777559983178381 seconds. Re-try count: 1/5: An error occurred (404) when calling the HeadObject operation: Not Found
s3 failed, retry after 4.387371754781057 seconds. Re-try count: 2/5: An error occurred (404) when calling the HeadObject operation: Not Found
s3 failed, retry after 9.891766225948565 seconds. Re-try count: 3/5: An error occurred (404) when calling the HeadObject operation: Not Found
etc.

a loop display.

Deepracer : Locally trained model not valid for import on DeepRacer console

Contenus pertinents