added missing file for evaluating the ensemble test predictions

import sys

import os

import numpy as np

import ast

from sklearn.metrics import f1_score,confusion_matrix,precision_score,recall_score

from scipy.stats import pearsonr

from eval_helpers import evaluate,extensive_evaluate,evaluate_regression,extensive_evaluate_regression

##script for evaluating final ensemble votings

S1=["changes_possession" ,"volition" ,"stationary" ,"location_of_event" ,"existed_before" ,"awareness" ,"exists_as_physical" ,"sentient" ,"existed_after" ,"existed_during" ,"makes_physical_contact" ,"destroyed" ,"change_of_location" ,"created" ,"instigation" ,"manipulated_by_another" ,"change_of_state", "predicate_changed_argument"]

S2=[

"awareness",

                "change_of_location",

                "change_of_possession",

                "change_of_state",

                "change_of_state_continuous",

                "existed_after",

                "existed_before",

                "existed_during",

                "instigation",

                "partitive",

                "sentient",

                "volition",

                "was_for_benefit",

                "was_used"]

S1=list(sorted(S1))

print(S1)

### these parameters must be set ###

#voter model ids

runids=list(range(0,50))

#are we doing regression or multi-label?

REGR = False

#SPR1:18 classes, SPR2:14

NCLASSES = 18

res={}

full = []

if NCLASSES == 14:

    S1=S2

def maxto1(x):

    #converts soft vote to hard vote

    for i in range(x.shape[0]):

        a=np.argmax(x[i])

        b=np.argmin(x[i])

        x[i][a] = 1

        x[i][b] = 0

    return None

def tolikert(x):

    #crops relu predictions to likert scale

    new=np.zeros(x.shape)

    for i in range(x.shape[0]):

        for j in range(x.shape[1]):

            if x[i][j] < 1:

                new[i][j] = 1

            elif x[i][j] > 5:

                new[i][j]=5

            else:

                new[i][j] = x[i][j]

    return new

N=0

for fi in os.listdir("bertbaselogs"):

    with open("bertbaselogs/"+fi,"r") as f:

        if int(fi.split("-")[0]) in runids:

            string=f.read()

            #get gold labels once from log file

            if N == 0:

                gold = [np.array(x) for x in ast.literal_eval(string.split("gold: ")[1].split("\n")[0])][:NCLASSES]

            #get test predictions from log file

            preds = [np.array(x) for x in ast.literal_eval(string.split("predictions: ")[1].split("\n")[0])][:NCLASSES]

            #probabilities to hard

            if not REGR:

                [maxto1(x) for x in preds]

            #relu scores to likert scale with simple crop 1 = scores < 1 ; 5 = scores > 5

            if REGR:

                preds=[tolikert(p) for p in preds]

            #if this is the first voter, intitialize committee predictions

            if not full:

                full = preds

            #else aggregate

            else:

                full = [full[i]+preds[i] for i in range(len(preds))]

            N+=1

            print(N,"voters loaded")

#print final results

if REGR:

    ext=extensive_evaluate_regression(gold,full,NCLASSES)

    print(list(zip(S1,ext[2].tolist())))

    print(evaluate_regression(gold,[tolikert(x/N) for x in full],NCLASSES))

else:

    ext=extensive_evaluate(gold,full,NCLASSES)

    print(list(zip(S1,ext[2].tolist())))

    print("mean F1","macro F1","micro F1")

    print(evaluate(gold,full,NCLASSES))