Yes I expect it to take some time! I have already setup a cluster to let it run over the weekend! I am very curious to see how HTM performs for longer steps ahead prediction. As expected, with "predictionSteps": [96] the swarm runs smoothly. I am changed the model_params manually to force a 96-vector output and also increase n as suggested here:
To go back to the original problem, it looks like during swarming it is trying to access the results of a model that is still running, any ideas how to alleviate that? I am trying to reduce the number of rows to swarm over but not sure if that will work…