Work in progress
To-do List
GeoServer LDAP integration
Dataverse LDAP integration
Check high memory allocation issue with CUDA/CuPy
The reason is identified: pre-allocation of GPU memory to speed-up processing.
Unfortunately this doesn’t work nicely with unified CPU-GPU memory architecture
Framework specific solutions exists and will be implemented.
Spark usage example
Waiting List
Application: Paraview - 3D data analysis and visualization application
Framework: Geonode - Open Source Geospatial Content Management System
Application: QGIS Server - WMS, WFS and WCS implementation that implements advanced cartographic features for thematic mapping
Application: Spyder - Scientific Python development environment
Finished
ODC installation
Application: ODK - Open Data Kit
Application : SNAP - SNAP and the Sentinel Toolboxes
Application: Firefox - Safe and easy web browser from Mozilla
Application: PyCharm - Python Integrated Development Environment (IDE)
Application: FragStats - calculates landscape metrics (can be installed through Wine)
Application: GMTSAR - InSAR processing system based on GMT
Application: Glueviz - Multi-dimensional linked-data exploration
Framework: Horovod - Framework for distributed deep learning training using TensorFlow, Keras, PyTorch, and Apache MXNet.
Framework: TensorFlow Addons - Useful extra functionality for TensorFlow 2.x maintained by SIG-addons
Framework: fast.ai - Making neural nets uncool again
Framework: SpaCy - Industrial-strength NLP
Language: Rust - Fast and memory-efficient programming language
Application: NetLogo - Multi-agent programmable modeling environment
Application: GRASS - Geographic Resources Analysis Support System
Framework: MLflow - Open source platform for the machine learning lifecycle
System Upgrade: Python 3.6 -> Python 3.8
Python 3.6 currently receives only security updates and has end-of-life of 2021/12.
NVIDIA L4T is based on Ubuntu 18.04, which has Python 3.6. Some packages are only available for Python 3.6 (e.g. tensorrt)
Alternative Docker files for Ubuntu 20.04 + Python 3.8 are in production.
All packages are migrated, except the following heavy ones:
TensorFlow: Build commands ready, to be tested.
MXNet: Build commands ready, to be tested.
ONNX Runtime: Build commands ready, to be tested.
Keras: Pending TensorFlow
Bug Fix: JAX CPU-backend
This was a difficult bug to identify, but it is fixed now (https://github.com/google/jax/issues/5679)
Will be available with Python 3.8
Application: Orange3 - Interactive data analysis
Built and tested, ready to deploy.
Application: RStudio - Integrated development environment for R
Built, but QT GLX fails at runtime.
It looks like we need to change VNC server again, possibly to TurboVNC
Tried TurboVNC, but solution also works with TigerVNC.
PostgreSQL LDAP integration