#!/usr/bin/python3
#-*-coding:utf-8-*-
#$File: model_eval.py
#$Date: Sat Jun 4 09:59:32 2016
#$Author: Like Ma <milkpku[at]gmail[dot]com>
import tensorflow as tf
import numpy as np
import sys
import argparse
class Evaluer(object):
def __init__(self, model_folder, checkpoint_file):
sys.path.append(model_folder)
from model import get_model
from dataset import load_data
self.dataset = load_data('validation')
self.sess = tf.InteractiveSession()
self.model = get_model('policy')
saver = tf.train.Saver()
saver.restore(self.sess, checkpoint_file)
def evalue_topN(self, batch, N=5):
data, label = self.dataset.next_batch(batch)
input_dict = {}
for var, subdata in zip(self.model.inputs, data):
input_dict[var] = subdata
pred = self.model.pred.eval(feed_dict=input_dict)
# accuracy
pred_flat = pred.reshape([-1, 9*10*16])
label_flat = label.reshape([-1, 9*10*16])
accuracy = np.mean(pred_flat.argmax(axis=1)==label_flat.argmax(axis=1))
# topN
topN = 0
for prob, ind in zip(pred_flat, label_flat.argmax(axis=1)):
v = prob[ind]
tmp = np.copy(prob)
tmp.sort()
if v >= tmp[-N]:
topN += 1
topN /= len(pred)
print('accuracy %.2f, top%d %.2f' % (accuracy, N, topN))
return data, label, pred
if __name__=='__main__':
parser = argparse.ArgumentParser()
parser.add_argument('model_folder', help='Foler for model')
parser.add_argument('checkpoint_file', help='File path of checkpoint')
args = parser.parse_args()
evaluer = Evaluer(args.model_folder, args.checkpoint_file)
#for i in range(10):
# evaluer.evalue_topN(100, 3)
data, label, pred = evaluer.evalue_topN(100,3)
import pickle
fh = open('pred.tensor', 'wb')
pickle.dump([data, label, pred], fh)
fh.close()
