Releases: awslabs/stable-diffusion-aws-extension
V1.5.0 extension package
The latest version primarily includes the following enhancements and innovations:
- Support webUI V1.8.0
- Support extension ReActor for Stable Diffusion
- Add API Debugger feature
- Support LoRa model training through Kohya_ss
- Improve customer experience on Resource management in webUI
- Optimize the method of customizing BYOC extensions, and support optimization startup
- Support auto-scale ability for real-time inference endpoints
- Support the customizable range for the numbers of instances during auto-scaling
- Optimize installation time by 67%, down to 4 minutes
- Upgrade Python version from 3.9 to 3.10
- Bugs fix
v1.4.0 extension package
The latest version primarily includes the following enhancements and innovations:
- Support webUI V1.7.0
- Support multiple Amazon stacks deployment in one account
- Enhance cold-start time of Amazon SageMaker inference time by 70%, down to 3 mins
- Support image generation through Amazon SageMaker real-time inference endpoint
- Improve customer experience on webUI, containing multiple level filters for tables of roles, users, models and inference endpoints, along with resource delete feature
- Bugs fix
v1.3.0 extension package
The latest version primarily includes the following enhancements and innovations:
- New Model Support: Stable Diffusion Turbo
- BYOC Feature: Enables customers to customize the WebUI plugin through BYOC (Bring Your Own Container).
- Multi-User Management: Supports user update without requiring to restart the WebUI
- Cloud-Only Edition: Allows users to update the WebUI to a version that exclusively supports cloud-based operation in client without GPU resource.
- API: Launches APIs that comply with the OpenAPI standard, facilitating easier querying and invocation by developers.
Additionally, the preview version includes:
- New Model Support: Stable Diffusion LCM.
- Diffusers: Offers inference capabilities based on the diffusers library, including direct invocation through both WebUI and API.
v1.2.1 extension package
Minor feature update and issue fix in list below:
- Create LambdaDeployRole in the shared stack to avoid issues caused by duplicate resources.
- Add validation for the S3 bucket name.
- Address the issue with Lambda quota.
- When two people are running the graph together (let's assume A and B), B's graph is based on the keywords written by A.
- After uploading the Lora model through the extension's web UI, it cannot be selected if the model is not available locally.
- Provide additional extra generate API support.
- After uploading the controlnet model through the extension's web UI, it cannot be selected if the model is not available locally.
- After uploading the embedding model through the extension's web UI, it cannot be used if the model is not available locally.
- Modify the scaling-down strategy to ensure that it is not triggered too early, which could result in longer cold start times.
- DDB (DynamoDB) is not promptly updated when removing SageMaker Endpoint from the web UI.
v1.2.0 extension package
- Version update support
- Multiple user management
- New model upload method (from URL to S3)
- webUI 1.6.0 support (XL + ControlNet, Refiner)
v1.1.0 extension package
- webUI 1.5.1, SDXL1.0 support
- VAE support
- Upload model from local to S3
- Multi-controlnet
- Endpoint auto scale
- New filter for inference Job
v1.0.1 extension package
- Upgrade the version support of webUI to V1.4.0
- Upgrade the version support of ControlNet to V1.1.227
- Upgrade the version support of Dreambooth to V1.0.14
- Support txt2img API Direct call
- Support full features in img2img except batch
- Support full features in ControlNet except multi-ControlNet
- Auto-refresh inference results in Output section
- Support filter in Inference Job ID
- Support rollback or update the middleware
- Bugs fix
v1.0.0 extension package
- Amazon SageMaker main tab with authentication, cloud assets management, dataset managerment
- txt2img/img2img inference integration
- Dreambooth/Controlnet inference & training integration