Release Note (0.2.0)
Release time: 2022-01-31 00:00:00
This is the first release of Aquila HUB includes improvements in performance, supports distribution of Fasttext models through IPFS and bug fixes.
New features: 1
Total features: 4
[Announcement] IPFS
support for Fasttext
models is added (refactor)
Aquila HUB will support IPFS to enable p2p model download (fasttext models only) from a CID (multaddress). This is subject to removal / re-implementation through a different approach as we test and make progress.
[Added] What’s new in Aquila HUB
Request batching is added to improve batched predictions. This allows AquilaHUB to make model predictions and to serve multiple client requests in a batch.
[Fixed] Bugs fixed
Several minor bugs are fixed and only the highlights are given below. To see all fixes, please refer to this release's commits.
- ed9563d Upgraded dependencies, removed conflicting disk writing, fixed unwanted model disk persistence per database.
- 8edfc70 Enabled backup & preloading from backup, improved conflict-free disk persistence.
[Deprecation Notice] What’s going to change
No planned deprecation
[Removed] What’s changed
No features removed
[Known Issues] We’re aware of these issues
- AquilaHUB is safe to deploy as a multi-instance. However, it doesn't support model autoloading from shared storage until restarted.