Cusignal nodes for Greenflow.

[avolkov1]

Added convolution nodes and a convolution notebook example.
Work in progress.

I'm going to add more examples.

Refer to notebook:
https://github.com/avolkov1/gQuant/blob/cusignal_example/gQuant/plugins/cusignal_plugin/notebooks/convolution_examples.ipynb

The plugin can be found here:
https://github.com/avolkov1/gQuant/tree/cusignal_example/gQuant/plugins/cusignal_plugin/greenflow_cusignal_plugin

I am currently using the "greenflowrc" to register the plugin. Later I'll add a setup.py and register via "greenflow.plugin".

To run checkout my branch. Then set "greenflowrc" as follows:

[ModuleFiles]
my_node = %(MODULEPATH)s/my_node.py
greenflow_cusignal_plugin = %(MODULEPATH)s/greenflow_cusignal_plugin

Where my modulepath is a directory from which I symlinked to the plugins.

$ ll modules
greenflow_cusignal_plugin -> <repo>/gQuant/plugins/cusignal_plugin/greenflow_cusignal_plugin
my_node.py -> <repo>/gQuant/plugins/gquant_plugin/modules/my_node.py

I startup jupyter lab via:

MODULEPATH=<repo>/workspace/modules \
GREENFLOW_CONFIG=<repo>/workspace/greenflowrc \
  jupyter lab --ip=0.0.0.0 --ContentsManager.allow_hidden=True --no-browser

I added pydot plotting options to "TaskGraph.draw". You'll need to pip install -e greenflow.

[yidong72]

I added a few comments. Since this is a stand alone plugin, could you add a README file to describe how to install and use this plugin like other plugins?

[yidong72]

good point. We need to change the API in the future to something like

def get_node(self, key):

and put a warning message inside this function that it is deprecated.

[avolkov1]

Yes, that makes sense. We can do this in the future.

[yidong72]

In the portfolio_trade.gq.yaml file.

- conf: {}
  id: ''
  inputs:
    in1: cumulative_return.cum_return
    in2: sharpe_ratio.sharpe_out
  module: rapids_modules
  type: Output_Collector

Need to remove the module: rapids_modules line.

Similarly in the simple_trade_gq.yaml file, need to remove the same line.

- conf: {}
  id: ''
  inputs:
    in1: stock_selector.stock_name
    in2: lineplot.lineplot
    in3: barplot.barplot
    in4: sharpe_ratio.sharpe_out
    in5: cumulative_return.cum_return
    in6: stock_data.cudf_out
  module: rapids_modules
  type: Output_Collector

also in the tutorial_intro.gq.yaml and get_return_feature.gq.yaml files.

[yidong72]

unittest tests/unit/test_performance.py failed, need to fix it.

(base) quant@21dc6bc7dbe0:~/greenflow/plugins/gquant_plugin$ python -m unittest tests/unit/test_performance.py
EE
======================================================================
ERROR: test_ports_setup_performance (tests.unit.test_performance.TestPerformance)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/quant/greenflow/plugins/gquant_plugin/tests/unit/test_performance.py", line 92, in test_ports_setup_performance
    stats.stats[key][0], self.ports_setup_ref[dict_key])
KeyError: 'ports_setup.output_collector_node.py'

======================================================================
ERROR: test_meta_setup_performance (tests.unit.test_performance.TestPerformance)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/quant/greenflow/plugins/gquant_plugin/tests/unit/test_performance.py", line 107, in test_meta_setup_performance
    stats.stats[key][0], self.meta_setup_ref[dict_key])
KeyError: 'meta_setup.output_collector_node.py'

----------------------------------------------------------------------
Ran 2 tests in 0.382s

[avolkov1]

Thanks for the review. I fixed the unit test to account for output collector calls and added a README file.

[yidong72]

looks good. I will merge it.