Skip to content

Custom model import from Amazon Bedrock#232

Merged
daisuke-awaji merged 1 commit intomainfrom
bedrock-custom-model-import
Sep 16, 2025
Merged

Custom model import from Amazon Bedrock#232
daisuke-awaji merged 1 commit intomainfrom
bedrock-custom-model-import

Conversation

@daisuke-awaji
Copy link
Contributor

@daisuke-awaji daisuke-awaji commented Aug 28, 2025

📚 Custom Model Import Configuration Guide

Overview

This PR adds comprehensive documentation for configuring custom models imported using Amazon Bedrock's Custom Model Import feature for use with Bedrock Engineer.

What's Added

📖 Documentation Files

  • English Guide: docs/custom-model-import/README.md
  • Japanese Guide: docs/custom-model-import/README-ja.md
  • README Updates: Added links to the new documentation in both English and Japanese README files

📋 Documentation Content

The guides cover:

  1. Custom Model Import Process

    • Model download from Hugging Face
    • tokenizer_config.json configuration for Converse API compatibility
    • S3 upload and import job execution
  2. Bedrock Engineer Configuration

    • Adding custom model definitions to models.ts
    • Required and optional parameters explanation
    • Regional availability settings
  3. Troubleshooting Section

    • Common errors and solutions
    • ModelNotReadyException handling
    • ARN validation issues
    • Streaming tool use configuration

Why This Change?

  • User Enablement: Allows users to leverage open-source models from Hugging Face within Bedrock Engineer
  • Comprehensive Support: Provides step-by-step instructions for the entire process
  • Error Prevention: Includes troubleshooting guide to help users avoid common pitfalls
  • Experimental Feature: Clearly marked as experimental with appropriate warnings

Example Use Case

The documentation uses DeepSeek-R1-Distill-Llama-8B as a practical example, showing:

  • Complete setup process from download to configuration
  • Specific chat_template modifications for Llama-3.1 format
  • Proper ARN configuration in Bedrock Engineer

Key Configuration Parameters

Parameter Description Example
baseId Complete model ARN arn:aws:bedrock:us-east-1:123456789012:imported-model/xxxxx
supportsStreamingToolUse Streaming support for tool usage false (recommended for most custom models)
availability Available regions { base: ['us-east-1'] }

Target Audience

  • Developers wanting to use custom/open-source models with Bedrock Engineer
  • Users interested in experimenting with models beyond the standard Bedrock offerings
  • Teams requiring specific model capabilities not available in standard models

Testing

  • Documentation formatting and readability verified
  • Code examples tested for accuracy
  • Links and references validated
  • Both English and Japanese versions synchronized

Notes

  • This feature is marked as Experimental and may undergo breaking changes
  • The documentation emphasizes proper ARN handling and streaming configuration
  • Includes comprehensive troubleshooting for common setup issues

- Add comprehensive Japanese documentation for custom model import configuration
- Add English version of custom model import documentation
- Update README.md and README-ja.md with links to new documentation
- Documentation includes:
  - DeepSeek-R1-Distill-Llama-8B import example
  - Complete tokenizer_config.json configuration
  - Bedrock Engineer models.ts setup instructions
  - Parameter details and troubleshooting guide
@daisuke-awaji daisuke-awaji merged commit ab2a333 into main Sep 16, 2025
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant