69 lines
2 KiB
Python
69 lines
2 KiB
Python
|
import numpy as np
|
||
|
|
from numpy import genfromtxt
|
||
|
|
import csv
|
||
|
|
import pandas
|
||
|
|
import argparse
|
||
|
|
|
||
|
|
def sample_predciton(path, rate):
|
||
|
|
data = pandas.read_csv(path).values
|
||
|
|
task_list = [0, 1, 2, 3, 4, 5, 6]
|
||
|
|
|
||
|
|
start = 0
|
||
|
|
stop = 0
|
||
|
|
num_unique = np.unique(data[:,1])
|
||
|
|
|
||
|
|
samples = []
|
||
|
|
for j in task_list:
|
||
|
|
for i in num_unique:
|
||
|
|
inx = np.where((data[:,1] == i) & (data[:,-2] == j))
|
||
|
|
samples.append(data[inx])
|
||
|
|
|
||
|
|
for i in range(len(samples)):
|
||
|
|
n = int(len(samples[i])*(100-rate)/100)
|
||
|
|
if n == 0:
|
||
|
|
n = 1
|
||
|
|
samples[i] = samples[i][:-n]
|
||
|
|
if len(samples[i]) == 0:
|
||
|
|
print('len of after sampling',len(samples[i]))
|
||
|
|
|
||
|
|
return np.vstack(samples)
|
||
|
|
|
||
|
|
def main():
|
||
|
|
# parsing parameters
|
||
|
|
parser = argparse.ArgumentParser(description='')
|
||
|
|
parser.add_argument('--batch_size', type=int, default=8, help='batch size')
|
||
|
|
parser.add_argument('--lr', type=float, default=1e-4, help='learning rate')
|
||
|
|
parser.add_argument('--hidden_size', type=int, default=128, help='hidden_size')
|
||
|
|
parser.add_argument('--model_type', type=str, default='lstmlast', help='model type')
|
||
|
|
|
||
|
|
args = parser.parse_args()
|
||
|
|
|
||
|
|
task = np.arange(7)
|
||
|
|
user_num = 5
|
||
|
|
bs = args.batch_size
|
||
|
|
lr = args.lr # 1e-4
|
||
|
|
hs = args.hidden_size #128
|
||
|
|
model_type = args.model_type #'lstmlast'
|
||
|
|
|
||
|
|
rate = [10, 20, 30, 40, 50, 60, 70, 80, 90]
|
||
|
|
|
||
|
|
for i in task:
|
||
|
|
for j in rate:
|
||
|
|
for l in range(user_num):
|
||
|
|
pred_path = "prediction/task" + str(i) + "/" + model_type + "_bs_" + str(bs) + "_lr_" + str(lr) + "_hidden_size_" + str(hs) + "/user" + str(l) + "_pred.csv"
|
||
|
|
save_path = "prediction/task" + str(i) + "/" + model_type + "_bs_" + str(bs) + "_lr_" + str(lr) + "_hidden_size_" + str(hs) + "/user" + str(l) + "_rate_" + str(j) + "_pred.csv"
|
||
|
|
data = sample_predciton(pred_path, j)
|
||
|
|
|
||
|
|
head = []
|
||
|
|
for r in range(7):
|
||
|
|
head.append('act'+str(r+1))
|
||
|
|
head.append('task_name')
|
||
|
|
head.append('gt')
|
||
|
|
head.insert(0,'action_id')
|
||
|
|
pandas.DataFrame(data[:,1:]).to_csv(save_path, header=head)
|
||
|
|
|
||
|
|
if __name__ == '__main__':
|
||
|
|
# split the prediction by length, from 10% to 90%
|
||
|
|
main()
|
||
|
|
|