ClientError for BatchTransform with BlazingText Classifier

I have successfully created the blazing text classifier according to the example code provided in aws-sagemaker examples. The model is deployed successfully and make predictions for new custom test cases. But I get the following error while attempting to perform BatchTransform with a file stored in S3.
My code sample for processing TestFile and training file is the same using nltk tokenizer. And after processing I am saving the test file for batch transform with the syntax:
"source": "source_sentence_0"
"source"; "source_sentence_1"

**CODE for batch transform:**
`testfile_location = 's3://{}/{}/batch/{}'.format(bucket, prefix, batch_file)
bt_transformer = bt_model.transformer(1, 'ml.m4.xlarge', strategy='SingleRecord', accept = "application/jsonlines", output_path = s3_output_location)
bt_transformer.transform(testfile_location, split_type='Line', content_type = "application/jsonlines")
bt_transformer.wait()`

Error message received is:
Customer Error: Unable to decode payload: Incorrect data format. (caused by ValueError)**

Training Sample:
__label__0 xperia xa ultra test results - sony mobile ( global english )
__label__0 san esteban de gormaz english tutors

File used for Batch Transform:
{"source": "fifiminoucha3 - Stardoll | English"}
{"source": "How do you write -9804 in English spelling ?"}

Start with the batch-transform code and the training and batch input samples shown in the issue, then compare them with the referenced aws-sagemaker examples. Reproduce the `Unable to decode payload` error using the S3 test file; done means BatchTransform accepts the JSON Lines input and produces predictions.

Written by the indexing model from the issue text.