#Release Notes

#AI 2.0.0

#New and Optimized Features

#NPU Support

By introducing the NPU Operator, NPU hardware management is supported, simplifying the setup for using NPUs and enhancing hardware acceleration.

#Queueing & Admission Control

By introducing the Alauda Build of Kueue, queue management and admission control are implemented, optimizing task scheduling and resource allocation.

#Model Registry

The integration of Kubeflow Model Registry enhances model management and version control, simplifying the process of model registration and storage.

#Distributed Inference

With the addition of llm-d, distributed inference is now supported, improving the performance and resource efficiency of large-scale inference tasks.

#Leader-Worker Set Support

Through the introduction of the Alauda Build of LeaderWorkerSet, distributed training task management is supported, simplifying task distribution and coordination in a Leader-Worker model.

#AI Gateway

The new Alauda build of Envoy AI Gateway optimizes traffic management and security, providing enhanced AI service proxy capabilities.

#Trainer Orchestration

By supporting Kubeflow Trainer v2, task scheduling and management for model training have been enhanced, enabling more flexible and efficient training workflows.

#Pipelines Orchestration

Integration with Kubeflow Pipeline allows for more efficient pipeline orchestration and task management, improving automation across workflows.

#Vector Storage

The introduction of Milvus provides an efficient vector storage solution, supporting large-scale vector data storage and fast retrieval.

#NVIDIA DRA

By introducing the NVIDIA DRA Driver for GPUs, GPU resource management and scheduling have been optimized, increasing GPU utilization and performance.

#Node Feature Discovery

The Alauda Build of Node Feature Discovery enables automatic discovery of hardware features and node labeling, improving flexibility in node resource management.

#Llama Stack Runtime

The introduction of Llama Stack enhances AI agent capabilities by providing an efficient runtime environment for supporting distributed AI tasks.

#Generative AI

By introducing Kserve's Generative AI section, support for generative AI has been enhanced, optimizing model deployment and inference, especially for generative AI applications.

#Low-Code Builder Integration

By updating the Dify version and providing deployable Charts, the low-code application building process has been simplified, further enhancing AI application development efficiency.

#Deprecated Features

#UI Fine-Tuning & Pretraining Deprecation

Due to the lack of general applicability, complex horizontal scaling, and the fact that this is not a mainstream method in the industry for model training and fine-tuning, this feature is deprecated. It is recommended to use notebook-based model fine-tuning and training instead.

#Secret Manage Deprecation

The Secret Manage feature is deprecated due to the lack of a valid use case, as manual GitLab integration is no longer required and no longer applicable.

#Fixed Issues

• When updating the inference service resource yaml through the page, the volumeMount field is missing, which can cause the inference service to fail to start properly
• In older versions, GraphQL queries (POST by default) were incorrectly intercepted by the gateway layer and checked for create permission. In the new version, requests sent to the /api/graphql interface are correctly treated as get read permissions by the RBAC interceptor, ensuring that users with read-only roles can read and access page content containing GraphQL data streams without problems.

#Known Issues

• After deleting a model, the list page fails to reflect the deletion result immediately, and the deleted model still briefly exists in the list. Temporary solution, manually refresh the page.
• Modifying library_name in Gitlab by directly editing the readme file does not synchronize the model type change on the page.
  Temporary solution: Use UI operation to modify the library_name to avoid direct operation in Gitlab.