I created the job using the base folder as the input path, something like python bin/comprehend-ssie-annotation-tool-cli.py --input-s3-path s3://{bucket}/docs/
...But the job failed due to NoSuchKey errors in the pre-annotation Lambda function:
[ERROR] NoSuchKey: An error occurred (NoSuchKey) when calling the GetObject operation: The specified key does not exist.
Traceback (most recent call last):
File "/var/task/lambdas/pre_human_task_lambda.py", line 346, in lambda_handler
pdf_s3_resp = s3_client.get_object_response_from_s3(source_ref)
File "/var/task/utils/s3_helper.py", line 36, in get_object_response_from_s3
return self.s3_client.get_object(Bucket=bucket, Key=path)
File "/var/task/botocore/client.py", line 514, in _api_call
return self._make_api_call(operation_name, kwargs)
File "/var/task/botocore/client.py", line 938, in _make_api_call
raise error_class(parsed_response, operation_name)
Looking at the generated input manifest, it seems like the preparation hasn't correctly handled the nested folders because everything is showing as filenames directly under the input-s3-path - for example:
I didn't see anything in the README guide suggesting we can't have nested folders in the input prefix, so not quite sure whether this is a bug or a feature request?
Today I tried to use the CLI to create a labelling job from an S3 prefix with nested folders - for example:
I created the job using the base folder as the input path, something like
python bin/comprehend-ssie-annotation-tool-cli.py --input-s3-path s3://{bucket}/docs/
...But the job failed due to
NoSuchKey
errors in the pre-annotation Lambda function:Looking at the generated input manifest, it seems like the preparation hasn't correctly handled the nested folders because everything is showing as filenames directly under the
input-s3-path
- for example:I didn't see anything in the README guide suggesting we can't have nested folders in the input prefix, so not quite sure whether this is a bug or a feature request?