search

NVlabs / FPSci

Aim Training Experiments
Other
70 stars 23 forks source link

Add a two interval forced choice sample #401

Closed jspjutNV closed 1 year ago

jspjutNV commented 1 year ago

This PR adds a sample of 2 interval forced choice. This sample should probably be updated once #392 is merged, but should work fine for earlier versions of FPSci.

bboudaoud-nv commented 1 year ago

I suggest we wait until merging the TaskFlow branch then make this example use tasks to implement the 2IFC model. This sample could still be useful, but I'd like to know more about why certain things are done (i.e., logToSingleDb = False is fairly uncommon, so I'd like to avoid providing samples w/ behavior that is atypical for FPSci configuration).

jspjutNV commented 1 year ago

logToSingleDb = False is fairly uncommon, so I'd like to avoid providing samples w/ behavior that is atypical for FPSci configuration

This came from the 1hit_sa2019 sample that I started this one from. There's probably a bunch of other stuff that could be dropped. I'll clean it up once we get to the task-level version. As is, it's still nice to have this around as an example of the old-style 2 interval forced choice design.

jspjutNV commented 1 year ago

76c97aa has a version using the pre-task-level configuration style of specification.

jspjutNV commented 1 year ago

As of 2279276 there appears to be an issue updating the status and the results db with this experiment, and I'm not sure what's going wrong. @bboudaoud-nv , the error I'm getting is the following (from log.txt):

User selected session: 60hz_delay. Updating now...
Warning insert rows with empty row value array ignored!
Warning insert rows with empty row value array ignored!
Warning insert rows with empty row value array ignored!
Warning insert rows with empty row value array ignored!
Creating table Experiments w/ SQL query:CREATE TABLE IF NOT EXISTS Experiments ( description TEXT NOT NULL ,time TEXT NOT NULL ,hash TEXT NOT NULL ,config TEXT NOT NULL );

Error in INSERT INTO statement (INSERT INTO Experiments VALUES('2IFC',): near "s": syntax error

The rest of the sql appears to succeed. I didn't go back and check if this was happening before I switched to task level. I suspect this isn't directly related to this experiment, but maybe it's a mistake in the experiment finding an issue we have a work around for.

jspjutNV commented 1 year ago

I removed parts of the scene config (left the name for easier adjustment for people who want to change the scene).

I also increased the target size slightly since it was a bit more difficult than the purpose of this sample justified.