38 posts tagged with "backend.ai"

The core engine of Backend.AI uses many open-source software, and it is itself being developed as an open source project. If you find any inconvenience or bugs while using Backend.AI, enterprise customers can use customer support and technical support channels for issue tracking and support, but it is also possible to contribute to the open-source directly. There are two ways to contribute: explaining in detail the issues or improvement ideas by creating an issue and directly contributing by fixing the code with a pull request. In this post, we introduce several things to keep in mind to communicate more effectively and quickly with the development team during the contribution process.

정말 작년말은 하던 일이 모두 잘되던 시기였다. 창업대회들도 많은 일들이 있었지만 결국 좋은 결과를 얻어냈고 소소하게 용돈벌이하려고 썼던 엘리스 스쿨튜터도 붙었다. 하지만 많은 일 중에서도 가장 기뻤던 건 고등학교 때부터 한번쯤 일해보고 싶었던 Lablup에 인턴으로 붙은 것이다. 약간 워커홀릭이라서 복학 전에 할 일들을 찾고 있었는데 이왕이면 내 성장에 도움이 되는 일들 하고 싶었다. 그때 마침 고등학교때부터 일해보고 싶었던 회사인 Lablup에서 인턴을 모집하는 공고를 보고 인턴을 지원하게 되었다.

Since the official release of Backend.AI 22.03 in April, additional improvements and bug fixes have been updated.

Backend.AI Core & UI (22.03)​

• Fixed an issue where the current user's login session information was not passed when running an app in wsproxy v2.
• Provided additional information on exception types so that client-side errors caused by HTTP 404 errors due to incorrect URLs and requesting non-existent data can be distinguished when API exceptions occur.
• Fixed an issue where folders that should be automatically connected (e.g., folders with names starting with ".") were missing if the user folder was not explicitly connected when creating a computation session.
• Changed the default value to use the PostgreSQL advisory lock used in the previous version, as there is a multi-threading issue with the gRPC library used in the new etcd-based distributed lock implementation for manager high availability configuration.
• Fixed a bug where statistics collection for a specific computation session was intermittently not performed correctly due to a race condition between container creation time and lifecycle event handlers.
• Added a UI that makes it easy to distinguish between Arm and x86-64 architectures in environments that use both.
• Improved File browser session to work on read-only folders.
• Fixed many UI typos, translations, and bugs related to internal behavior and settings.

Computation environment support added and improved​

• Added TensorFlow 2.9 environment
• Optimized JAX/Flax

Forklift (beta)​

Forklift, which helps users build arbitrary container images for use in Backend.AI computation sessions, is under development and will be released in beta in mid-June. You can use the additional container image build feature not only through the web UI but also through the API.

• Forklift exclusive web UI support
• Monitoring of requested image build status for each user
• Function to automatically insert arbitrary package installation commands (apt, pip, conda, etc.)
• Provides automatic Dockerfile creation and preview after specifying build options through the UI.
• Provides detailed console logs for each build step.
• Option to download the built image or automatically push it to the specified registry (on/off).

More features are coming soon!