We didn't experience the expected performance improvements. Locally under
unit test there was aboout an order of magnitude throughput increase, but
when run against the history benchmark, this new approach was 60% slower.
We're reverting this for now to perform further analysis and better
understand the performance profiling of the python history stack.
- Adds log to display the python version the algorithm is using.
- Fixes python algorithms that were failing because of small subtleties
like leading zeroes.
- Updates pythonnet with a version compiled with python 3.6 flags
- Changes in DockerfileFoundation: we now use miniconda to manage the python
environment.
- Took the opportunity to add NTLK (#1349), Tensorforce (#1369) and
PyTorch/Pyro (#1385).
- Changes readme in Algorithm.Python to show steps to install miniconda
In python algorithm using Universe Selection, the selector method should return a List<Symbol>. To make it more pythonic, we allow returning python list. The conversion is, then, performed in C# side.
Some python algorithms suffered corrections to run under the new python framework (pythonnet).
Others were deleted because some features will be supported in futures implementations.
Adds a method in AlgorithmPythonUtil to transform C# DateTime into Python datetime
Michael Handschuh