The Multi-Action Cascade Model of Conversation
MACMInitialization file creates the data files requried for the MACM model to run and puts them in the init_data directory.
Following is the way to execute the MACMInitialization.py from the root directory:
usage: MACMInitialization.py [-h] [-d DEVICEID]
event_file shocks_file time_min time_max
positional arguments:
event_file event file to be used to infer endo/exo-genous
influence.
shocks_file event file to be used to infer endo/exo-genous
influence.
time_min Start of training time.
time_max End of training time.
optional arguments:
-h, --help show this help message and exit
-d DEVICEID, --DeviceID DEVICEID
Device ID
Example (running on GPU device 1 ):
python MACM/MACMInitialization.py '/home/social-sim/SSDATA/CP4_Final/test_harness_1_macm_training_stance_no_dupes.csv' '/home/social-sim/MACMWorking/MACM/init_data/all_exogenous_shocks_cp4_scen1.csv' '2019-01-15T00:00:00Z' '2019-01-17T00:00:00Z' -d 1
The model could be run using the Run.py or by creating your own running program which uses the class file MACM.
Following is the way to execute the Run.py from the root directory:
usage: Run.py [-h] [-q] [--device-id DEVICE_ID] [-m]
START_TIME TICKS_TO_SIMULATE MAX_MEMORY_DEPTH
MEMORY_DEPTH_FACTOR
positional arguments:
START_TIME Start time of simulation.
TICKS_TO_SIMULATE Number of hours to run simulation.
MAX_MEMORY_DEPTH Max memory depth parameter.
MEMORY_DEPTH_FACTOR Memory depth factor parameter.
optional arguments:
-h, --help show this help message and exit
-q, --quiet Set for detailed output.
--device-id DEVICE_ID
CUDA device id.
-m, --dump_agent_memory
Dump received information, actionable information, and
attention span data. Considerably slows down model
runs.
Example (running on GPU device 0):
python Run.py '2019-02-01T00:00:00Z' 3 10 0.8 --device-id 0