Amazon Bedrock Nanonets-OCR Model Import and Inference Guide

Prerequisites

AWS account with Bedrock access
Nanonets-OCR model files stored in an S3 bucket
AWS CLI or Console access
boto3 Python library installed (pip install boto3)

Part 1: Create IAM Service Role

Before importing the model, you need to create an IAM service role that allows Bedrock to access your S3 bucket.

Step 1.1: Create the Service Role

Navigate to IAM Console: https://console.aws.amazon.com/iam/
Go to Roles → Create role
Select AWS service as trusted entity type
Choose Bedrock from the service list
Click Next

Step 1.2: Add Permissions Policy

Create a custom policy with the following permissions:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "BedrockS3Access",
            "Effect": "Allow",
            "Action": [
                "s3:GetObject",
                "s3:ListBucket"
            ],
            "Resource": [
                "arn:aws:s3:::<YOUR_BUCKET_NAME>",
                "arn:aws:s3:::<YOUR_BUCKET_NAME>/*"
            ],
            "Condition": {
                "StringEquals": {
                    "aws:ResourceAccount": "<YOUR_AWS_ACCOUNT_ID>"
                }
            }
        }
    ]
}

Replace:

<YOUR_BUCKET_NAME>: Your S3 bucket name (e.g., my-models-bucket)
<YOUR_AWS_ACCOUNT_ID>: Your AWS account ID (e.g., 123456789012)

Step 1.3: Configure Trust Relationship

Edit the trust relationship to include:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "BedrockAssumeRole",
            "Effect": "Allow",
            "Principal": {
                "Service": "bedrock.amazonaws.com"
            },
            "Action": "sts:AssumeRole",
            "Condition": {
                "StringEquals": {
                    "aws:SourceAccount": "<YOUR_AWS_ACCOUNT_ID>"
                },
                "ArnEquals": {
                    "aws:SourceArn": "arn:aws:bedrock:<YOUR_REGION>:<YOUR_AWS_ACCOUNT_ID>:model-import-job/*"
                }
            }
        }
    ]
}

Replace:

<YOUR_AWS_ACCOUNT_ID>: Your AWS account ID
<YOUR_REGION>: Your AWS region (e.g., us-west-2)

Step 1.4: Name and Create Role

Name the role (e.g., BedrockModelImportRole)
Add a description
Click Create role
Note down the Role ARN for later use

Part 2: Import Nanonets-OCR Model to Bedrock

Step 2.1: Navigate to Import Models Page

Go to Amazon Bedrock Import Models page:

https://<YOUR_REGION>.console.aws.amazon.com/bedrock/home?region=<YOUR_REGION>#/import-models

Replace <YOUR_REGION> with your region (e.g., us-west-2)

Step 2.2: Start Model Import

Click Import model button

Step 2.3: Configure Model Details

Model name: Enter a descriptive name for your model
- Example: nanonets-ocr-v1
Model import settings:
- Model import source: Select Amazon S3 bucket
- S3 URI: Enter the path to your model files
  s3://<YOUR_BUCKET_NAME>/<MODEL_FOLDER_PATH>/
  - Example: s3://my-models-bucket/nanonets-ocr-model/

Step 2.4: Configure Service Access

Under Service access, select Use existing service role
Choose the IAM role created in Part 1 (e.g., BedrockModelImportRole)

Step 2.5: Import the Model

Review all settings
Click Import model button

Step 2.6: Monitor Import Job

The import job will be created and take approximately 10-15 minutes to complete
You can monitor the progress on the Import Models page
The status will change from “In Progress” to “Completed”

Step 2.7: Locate Your Imported Model

Once completed, go to the Imported models tab:

https://<YOUR_REGION>.console.aws.amazon.com/bedrock/home?region=<YOUR_REGION>#/import-models

Click on your model name to view details
Note down the Model ARN - you’ll need this for inference
- Format: arn:aws:bedrock:<YOUR_REGION>:<YOUR_ACCOUNT_ID>:imported-model/<MODEL_ID>

Part 3: Use Model for Inference

Step 3.1: Set Up Python Environment

Install required dependencies:

pip install boto3

Step 3.2: Configure AWS Credentials

Ensure your AWS credentials are configured:

aws configure

Or set environment variables:

export AWS_ACCESS_KEY_ID=<YOUR_ACCESS_KEY>
export AWS_SECRET_ACCESS_KEY=<YOUR_SECRET_KEY>
export AWS_DEFAULT_REGION=<YOUR_REGION>

Step 3.3: Create Inference Script

Create a Python file bedrock_inference.py:

import boto3
import json
import base64
import time
from pathlib import Path

# Initialize Bedrock client
bedrock = boto3.client("bedrock-runtime", region_name="<YOUR_REGION>")

# Your imported model ARN
MODEL_ID = "arn:aws:bedrock:<YOUR_REGION>:<YOUR_ACCOUNT_ID>:imported-model/<MODEL_ID>"

# Load image (for vision models)
image_path = "<PATH_TO_YOUR_IMAGE>"
image_b64 = base64.b64encode(Path(image_path).read_bytes()).decode()

# Prepare request body
body = json.dumps({
    "messages": [{
        "role": "user",
        "content": [
            {
                "type": "image_url",
                "image_url": {
                    "url": f"data:image/jpeg;base64,{image_b64}"
                }
            },
            {
                "type": "text",
                "text": "Extract all information from this invoice as clean markdown."
            }
        ]
    }],
    "max_tokens": 20000,
    "temperature": 0.7
})

# Make inference request
start_time = time.time()
response = bedrock.invoke_model(
    modelId=MODEL_ID,
    body=body,
    contentType="application/json",
    accept="application/json"
)
elapsed_time = time.time() - start_time

# Parse response
result = json.loads(response["body"].read())

# Print results
print(result["choices"][0]["message"]["content"])
print(f"\n⏱️  Response time: {elapsed_time:.2f}s")

Replace:

<YOUR_REGION>: Your AWS region (e.g., us-west-2)
<YOUR_ACCOUNT_ID>: Your AWS account ID
<MODEL_ID>: Your imported model ID
<PATH_TO_YOUR_IMAGE>: Path to your test image

Step 3.4: Run Inference

python bedrock_inference.py

Required IAM Permissions for Inference

The user/role running inference needs:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "bedrock:InvokeModel"
            ],
            "Resource": "arn:aws:bedrock:<YOUR_REGION>:<YOUR_ACCOUNT_ID>:imported-model/*"
        }
    ]
}

Notes

Model import typically takes 10-15 minutes
Imported models can be deleted from the Bedrock console if no longer needed
You are charged for model storage and inference requests
Model inference latency depends on model size and complexity

Getting Started

Resources

Amazon Bedrock Nanonets-OCR Model Import and Inference Guide

Prerequisites

Part 1: Create IAM Service Role

Step 1.1: Create the Service Role

Step 1.2: Add Permissions Policy

Step 1.3: Configure Trust Relationship

Step 1.4: Name and Create Role

Part 2: Import Nanonets-OCR Model to Bedrock

Step 2.1: Navigate to Import Models Page

Step 2.2: Start Model Import

Step 2.3: Configure Model Details

Step 2.4: Configure Service Access

Step 2.5: Import the Model

Step 2.6: Monitor Import Job

Step 2.7: Locate Your Imported Model

Part 3: Use Model for Inference

Step 3.1: Set Up Python Environment

Step 3.2: Configure AWS Credentials

Step 3.3: Create Inference Script

Step 3.4: Run Inference

Required IAM Permissions for Inference

Notes

Additional Resources

Getting Started

Resources

​Prerequisites

​Part 1: Create IAM Service Role

​Step 1.1: Create the Service Role

​Step 1.2: Add Permissions Policy

​Step 1.3: Configure Trust Relationship

​Step 1.4: Name and Create Role

​Part 2: Import Nanonets-OCR Model to Bedrock

​Step 2.1: Navigate to Import Models Page

​Step 2.2: Start Model Import

​Step 2.3: Configure Model Details

​Step 2.4: Configure Service Access

​Step 2.5: Import the Model

​Step 2.6: Monitor Import Job

​Step 2.7: Locate Your Imported Model

​Part 3: Use Model for Inference

​Step 3.1: Set Up Python Environment

​Step 3.2: Configure AWS Credentials

​Step 3.3: Create Inference Script

​Step 3.4: Run Inference

​Required IAM Permissions for Inference

​Notes

​Additional Resources

Prerequisites

Part 1: Create IAM Service Role

Step 1.1: Create the Service Role

Step 1.2: Add Permissions Policy

Step 1.3: Configure Trust Relationship

Step 1.4: Name and Create Role

Part 2: Import Nanonets-OCR Model to Bedrock

Step 2.1: Navigate to Import Models Page

Step 2.2: Start Model Import

Step 2.3: Configure Model Details

Step 2.4: Configure Service Access

Step 2.5: Import the Model

Step 2.6: Monitor Import Job

Step 2.7: Locate Your Imported Model

Part 3: Use Model for Inference

Step 3.1: Set Up Python Environment

Step 3.2: Configure AWS Credentials

Step 3.3: Create Inference Script

Step 3.4: Run Inference

Required IAM Permissions for Inference

Notes

Additional Resources