updated endpoints

[surcyf123]

new endpoints:
urls = [
"http://172.218.204.83:2701",
"http://37.27.2.44:60102",
"http://184.67.78.114:42098",
]
I also won't be normalizing the scores within the reward endpoint anymore. We can always iterate through the scores in the csv and append a normalized scores after the fact. Will link a commit with updated reward endpoint response format once it is completed

endpoint response format is as follows:
{'Bert': -0.029786773025989532, 'DPO': -11.46875, 'MPNet': 0.029283182695508003, 'RLHF': -2.9505324363708496, 'Reciprocate': -5.0703125}

[surcyf123]

changed again to

    "http://142.182.6.112:55469", "http://142.182.6.112:55467", "http://142.182.6.112:55430", "http://142.182.6.112:55430", #"vast2",
    "http://184.67.78.114:42036", "http://184.67.78.114:42091", "http://184.67.78.114:42082", "http://184.67.78.114:42093", # "vast3",
    "http://37.27.2.44:60113", "http://37.27.2.44:60180", "http://37.27.2.44:60151", "http://37.27.2.44:60181", # "vast4",
]

fixed normalization to be a constant and not a moving average, so output format is reverted to

[{'Bert': [-0.014279717579483986, 1.0], 'DPO': [-14.8203125, 0.0726899802684784], 'MPNet': [0.5161169767379761, 1.0], 'RLHF': [5.964920997619629, 0.9999698400497437], 'Reciprocate': [6.05078125, 0.8049955368041992], 'Total Reward': 0.6632935911417007}, {'Bert': [-0.014279717579483986, 1.0], 'DPO': [-14.8203125, 0.0726899802684784], 'MPNet': [0.5161169767379761, 1.0], 'RLHF': [5.964920997619629, 0.9999698400497437], 'Reciprocate': [6.05078125, 0.8049955368041992], 'Total Reward': 0.6632935911417007}]

[surcyf123]

this should help us avoid postporcessing on the testing data as well.

[surcyf123]

we need a plan of a place where I should keep the updated endpoints so you know where to look if they changed.

[gee842]

hm so from my experience the best practice for this would be to have another flask server which will have an endpoint that returns the endpoints, which the testing code can directly call before running any experiments