Feature/intervention

[van-dai-vuong]

Description

This PR introduces the intervention feature to model and skf. Users can make interventions in two ways: (1) add interventions (at single or multiple time steps) to mean and variance for a specific hidden state directly, and (2) create an extra hidden state for intervention under Intervention component, then specify the hidden state which will be added the intervention.

Note: at the moment, when using (2), should add one intervention hidden state per intervention event.

Changes Made

model.filter, model.forecast, skf.filter, Data_process

Related Issue(s)

NA.

Checklist

• I have followed the project's coding conventions and style guidelines.
• I have updated or added documentation where necessary.
• I have rebased my branch onto the latest main (or relevant) branch.
• I have verified that no sensitive data, plots, or files are included.
• I have run all previous examples successfully, with no errors.
• I have tested my changes locally using the following commands/tests:

Notes for Reviewers

pytest

Run:

• python -m examples.intervention_direct for testing model: multiple interventions, directly modify hidden states.
• python -m examples.intervention_add_state for testing model: a single intervention, add intervention hidden state.
• python -m examples.skf_intervention_direct for testing skf.
• python -m examples.skf_intervention_add_statefor testing skf.

[zhanwen-xin]

All good from my side

[MichelWU99]

Hi,

When I try to launch the example I got a return of a bug :

Traceback (most recent call last):
File "/Users/michelwu/miniconda3/envs/canari/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/Users/michelwu/miniconda3/envs/canari/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/Users/michelwu/new_canari/examples/intervention_add_state.py", line 83, in
(mu_validation_preds, std_validation_preds, states) = model.lstm_train(
File "/Users/michelwu/new_canari/src/canari/model.py", line 1877, in lstm_train
self.smoother()
File "/Users/michelwu/new_canari/src/canari/model.py", line 1807, in smoother
self.lstm_states_history[0] = self.lstm_net.get_lstm_states(
TypeError: Sequential.get_lstm_states() takes 1 positional argument but 2 were given

Is it only from my side ?

[van-dai-vuong]

Hi,

When I try to launch the example I got a return of a bug :

Traceback (most recent call last): File "/Users/michelwu/miniconda3/envs/canari/lib/python3.10/runpy.py", line 196, in _run_module_as_main return _run_code(code, main_globals, None, File "/Users/michelwu/miniconda3/envs/canari/lib/python3.10/runpy.py", line 86, in _run_code exec(code, run_globals) File "/Users/michelwu/new_canari/examples/intervention_add_state.py", line 83, in (mu_validation_preds, std_validation_preds, states) = model.lstm_train( File "/Users/michelwu/new_canari/src/canari/model.py", line 1877, in lstm_train self.smoother() File "/Users/michelwu/new_canari/src/canari/model.py", line 1807, in smoother self.lstm_states_history[0] = self.lstm_net.get_lstm_states( TypeError: Sequential.get_lstm_states() takes 1 positional argument but 2 were given

Is it only from my side ?

@MichelWU99 I think you need to uninstall and install pytagi again. Do pip uninstall pytagi, then pip install pytagi. I checked, it runs on the server as well.

[MichelWU99]

All good from my side after uninstalling and installing pytagi