Hi All,
I am trying to host a ray actor within a flask app for deployment as an API. However, while it works fine on my local, something is not working when I deploy it to a server - it is taking more than 10 mintues to get a response.
My current code is getting stuck when calling compute_action. Is there some way to get ray to log more information to a file somewhere? Perhaps that would give some insight as to what it’s doing when you call compute_action. Right now it’s all pretty opaque – I know that I’m calling a ray function, and that as a result it’s likely to be marshalling the data that you’re passing into that function, sending it to the server processes that init started, they are (hopefully) receiving it and running the model, and then the results are being marshalled and passed back to the Flask process from the servers, then unmarshalled and returned to my Flask code. Somewhere in there, something is taking a long time to run, but without more logging it’s likely to be hard to pin down exactly what.
Any suggestions for how to see how this process is developing? It is being hosted on pythonanywhere, but if anyone has a better suggestion it would be much appreciated!