Receive Fine tuning format error when trying to fine tune using the console

The error you're encountering is likely due to formatting issues in your JSONL file. Looking at the example you provided, there are a few problems:

1. The file appears to contain comments (the line starting with //), which are not valid in JSONL format.

2. There's a syntax error in your JSON objects - in both examples, you're missing a quotation mark before "role" in the assistant's response. It should be {"role": "assistant"... instead of {role": "assistant"....

3. JSONL requires each JSON object to be on a single line with no additional new lines or empty lines.

For fine-tuning models in Amazon Bedrock, each line must be a valid JSON object. Make sure:

• Remove any comments or header lines
• Fix the missing quotation marks
• Ensure each JSON object is complete and properly formatted
• Verify there are no additional new lines between records

The JSONL file should look like this (with each JSON object on a single line):

{"schemaVersion": "bedrock-conversation-2024", "system": [{"text": "You are a digital assistant with a friendly personality"}], "messages": [{"role": "user", "content": [{"text": "What is the first step I must take before proceeding to change any information?"}]}, {"role": "assistant", "content": [{"text": "You must coordinate with all those who need access."}]}]}
{"schemaVersion": "bedrock-conversation-2024", "system": [{"text": "You are a digital assistant with a friendly personality"}], "messages": [{"role": "user", "content": [{"text": "Before I dive into tweaking details, what is the first step I need to take?"}]}, {"role": "assistant", "content": [{"text": "You must coordinate with all those who need access."}]}]}

Even if online JSONL validators accept your file, Amazon Bedrock has specific formatting requirements for fine-tuning jobs.
Sources
Troubleshooting model customization issues - Amazon Bedrock
Prepare your training datasets for fine-tuning and continued pre-training - Amazon Bedrock