lemonade-sdk/lemonade v8.0.0

on GitHub

Headline

Lemonade SDK v8 is a major release with the following new features:

Ubuntu support (@ramkrishna2910)

Ubuntu users now have access to:

• Vulkan acceleration for GGUF models on GPU.
• ONNX and GGUF models on CPU.
• Lemonade Server web app.

Support for any GGUF model on Hugging Face (@jeremyfowers)

1. Open the new model manager: run the server, open http://localhost:8000 in your browser, and click Model Management.
2. Use the form to register the model with Lemonade Server and install it.

Also available programmatically via the pull endpoint and CLI command.

Lemonade Server is a Tray App on Windows (@danielholanda)

On Windows, Lemonade Server will launch as a system tray app:

• Native look and feel, with the terminal log no longer showing by default.
• Right click the lemon icon to access handy utilities (load model, change port, open the web app, view logs, update Lemonade Server, etc.).
• Lemonade Server can now run automatically on startup.

Performance Overhaul (@jeremyfowers, @danielholanda)

• Lemonade_Server_Installer.exe with Hybrid Execution installs in under 2 minutes and takes under 1 GB of space.
• Lemonade Server starts in under 2 second.
• lemonade -h runs in under 1 second.

Breaking Changes

Lemonade Server:

• The load endpoint can no longer load arbitrary checkpoints, and can only load registered models by their model name.
  • Instead, the pull endpoint can now register and install arbitrary checkpoints under a new model name.
  • See the load and pull changes in https://lemonade-server.ai/docs/server/server_spec/ for details.
• lemonade serve, which was deprecated, has been removed. Use lemonade-server-dev serve instead.

Additional Improvements

• Added a delete endpoint, CLI command, and webapp button to Lemonade Server, to delete installed models from disk (@ramkrishna2910).
• The models in the webapp's model manager have colorful labels (@danielholanda).

Documentation

• Supported configurations (software, hardware, and operating systems) table added to the top-level README.md (@jeremyfowers, @ramkrishna2910, @danielholanda).
• Completely overhauled and improved Open WebUI featured app guide (@jeremyfowers, @vgodsoe).
• New featured app guide for Microsoft AI Dev Gallery (@danielholanda).
• General documentation improvements (@vgodsoe, @jeremyfowers).

Bugs Fixed

• Stopped Lemonade_Server_Installer.exe from blocking Hybrid Execution installation on PCs that block Windows registry reads (@ramkrishna2910, @amd-pworfolk).
• Fixed llamacpp tools in the CLI (@ramkrishna2910).
• Administrator privledges are no longer required to run Lemonade_Server_Installer.exe (@danielholanda).

• Adding processor check override for internal usage by @CordellBlanchard in #14
• Fix UnicodeDecodeError: 'utf-8' codec by @Looong01 in #22

New Contributors

• @CordellBlanchard made their first contribution in #14
• @Looong01 made their first contribution in #22

Full Changelog: v7.0.4...v8.0.0