search

AllenInstitute / mtrain_regimens

Repository to track training stages and regimens for visual behavior
Other
2 stars 4 forks source link

fix agent_socket bug #102

Closed mochic closed 4 years ago

mochic commented 4 years ago

We found a parameter bug.

mochic commented 4 years ago

@jeromelecoq Hey, I'm really sorry this keeps coming up but we keep hitting small-ish bugs. I'm confident that we are getting very close to something that won't have any more bugs.

jeromelecoq commented 4 years ago

what is this agent_socket: "127.0.0.1:5000 parameter doing? Are you overwriting which server it is talking to?

mochic commented 4 years ago

@jeromelecoq It's a hack we need to employ to polyfill a feature that Ross can hopefully add in later. It points to localhost at port 5000 which should be the agent socket. We use the agent socket to run the replay so that we can use agent's session setup/teardown to submit our behavior output to lims. I put it in a parameter rather than hardcoding it incase of some weird situation where we would need to change it. Port 5000 is the default agent socket for remote method requests.

jeromelecoq commented 4 years ago

@jeromelecoq It's a hack we need to employ to polyfill a feature that Ross can hopefully add in later. It points to localhost at port 5000 which should be the agent socket. We use the agent socket to run the replay so that we can use agent's session setup/teardown to submit our behavior output to lims. I put it in a parameter rather than hardcoding it incase of some weird situation where we would need to change it. Port 5000 is the default agent socket for remote method requests.

Sorry can you explain differently? Why do you need this agent to replay a script?

mochic commented 4 years ago

so that agent can perform the upload to lims