-
Notifications
You must be signed in to change notification settings - Fork 0
/
predict.py
197 lines (170 loc) · 8 KB
/
predict.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
import argparse
import os
import pickle
import random
import sys
import collections
import numpy as np
import tensorflow as tf
from keras import Model
from utils.myutils import batch_gen, init_tf, seq2sent
import keras
import keras.backend as K
from utils.model import create_model
from timeit import default_timer as timer
from models.custom.graphlayer import GCNLayer
import pickle
import json
import heapq
class NDArrayEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj, np.ndarray):
return obj.tolist()
return json.JSONEncoder.default(self, obj)
def gen_pred(fid_set, model, data, comstok, smltok, comlen, batchsize, config, strat='greedy'):
# right now, only greedy search is supported...
tdats, coms, wsmlnodes, wedge_1 = zip(*data.values())
tdats = np.array(tdats)
coms = np.array(coms)
wsmlnodes = np.array(wsmlnodes)
wedge_1 = np.array(wedge_1)
final_dict = collections.defaultdict(dict)
dict1 = {}
dict1["code sequence_token"] = tdats
dict1["code sequence"] = seq2sent(tdats[0], smltok)
dict1["graph node_token"] = wsmlnodes
dict1["graph node"] = seq2sent(wsmlnodes[0], smltok)
dict1["graph edge"] = wedge_1
dict1["topK"] = {}
dict1['topK_confidence_score'] = {}
dict1["topK_word"] = {}
outfn = "ICPC2020_GNN/modelout/predictions/layer_output.txt"
outf = open(outfn, 'w')
output_dict = collections.defaultdict(dict)
for i in range(1, comlen):
output_dict[i] = {}
results = model.predict([tdats, coms, wsmlnodes, wedge_1],
batch_size=batchsize)
att1 = Model(inputs=model.input, outputs=model.get_layer('dot_1').output)
att2 = Model(inputs=model.input, outputs=model.get_layer('dot_3').output)
emd1 = Model(inputs=model.input, outputs=model.get_layer('embedding_1').get_output_at(0))
emd2 = Model(inputs=model.input, outputs=model.get_layer('embedding_1').get_output_at(1))
att1_output = att1.predict([tdats, coms, wsmlnodes, wedge_1])
att2_output = att2.predict([tdats, coms, wsmlnodes, wedge_1])
emd1_output = emd1.predict([tdats, coms, wsmlnodes, wedge_1])
emd2_output = emd2.predict([tdats, coms, wsmlnodes, wedge_1])
outf.write("iteration{}\n".format(i))
outf.write("{}\t{}\t{}\n".format("Attention weight between code sequence and predicted document", att1_output.shape, att1_output))
outf.write("{}\t{}\t{}\n".format("Attention weight between graph node sequence and predicted document", att2_output.shape, att2_output))
outf.write("{}\t{}\t{}\n".format("Embedding for code sequence", emd1_output.shape, emd1_output))
outf.write("{}\t{}\t{}\n".format("Embedding for code graph node", emd2_output.shape, emd2_output))
# print(att1_output)
# print(att2_output)
# print(emd1_output)
output_dict[i]["Attention weight between code sequence and predicted document"] = att1_output
output_dict[i]["Attention weight between graph node sequence and predicted document"] = att2_output
output_dict[i]["Embedding for code sequence"] = emd1_output
output_dict[i]["Embedding for code graph node"] = emd2_output
for c, s in enumerate(results):
coms[c][i] = np.argmax(s)
dict1['topK_confidence_score'][i] = list(heapq.nlargest(10, [np.float(tmp) for tmp in s]))
dict1["topK"][i] = s.argsort()[-10:][::-1]
dict1["topK_word"][i] = seq2sent(dict1["topK"][i], comstok)
print("com_i", coms[c][i])
with open('ICPC2020_GNN/modelout/predictions/layer_output.json', 'w') as fp:
# pickle.dump(output_dict, fp)
fp.write(json.dumps(output_dict, cls=NDArrayEncoder))
with open('ICPC2020_GNN/modelout/predictions/input.json', 'w') as fp:
# pickle.dump(dict1, fp)
fp.write(json.dumps(dict1, cls=NDArrayEncoder))
outf.close()
# with open('ICPC2020_GNN/modelout/predictions/layer_output.json', 'rb') as fp:
# data = pickle.load(fp)
# print(data[2])
final_data = {}
for fid, com in zip(data.keys(), coms):
final_data[fid] = seq2sent(com, comstok)
return final_data, output_dict, dict1
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='')
parser.add_argument('model', type=str, default=None)
parser.add_argument('--modeltype', dest='modeltype', type=str, default=None)
parser.add_argument('--gpu', dest='gpu', type=str, default='')
parser.add_argument('--data', dest='dataprep', type=str, default='../data')
parser.add_argument('--outdir', dest='outdir', type=str, default='ICPC2020_GNN/modelout/')
parser.add_argument('--batch-size', dest='batchsize', type=int, default=1)
parser.add_argument('--outfile', dest='outfile', type=str, default=None)
args = parser.parse_args()
modelfile = args.model
outdir = args.outdir
dataprep = args.dataprep
gpu = args.gpu
batchsize = args.batchsize
modeltype = args.modeltype
outfile = args.outfile
config = dict()
# User set parameters#
config['maxastnodes'] = 100
config['asthops'] = 10
if modeltype == None:
modeltype = modelfile.split('_')[0].split('/')[-1]
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = gpu
tdatstok = pickle.load(open('%s/tdats.tok' % (dataprep), 'rb'), encoding='UTF-8')
comstok = pickle.load(open('%s/coms.tok' % (dataprep), 'rb'), encoding='UTF-8')
smltok = pickle.load(open('%s/smls.tok' % (dataprep), 'rb'), encoding='UTF-8')
seqdata = pickle.load(open('%s/dataset.pkl' % (dataprep), 'rb'))
allfids = list(seqdata['ctest'].keys())
datvocabsize = tdatstok.vocab_size
comvocabsize = comstok.vocab_size
smlvocabsize = smltok.vocab_size
config['tdatvocabsize'] = datvocabsize
config['comvocabsize'] = comvocabsize
config['smlvocabsize'] = smlvocabsize
# set sequence lengths
config['tdatlen'] = 50
config['comlen'] = len(list(seqdata['ctrain'].values())[0])
config['smllen'] = len(list(seqdata['strain_nodes'].values())[0])
config['batch_size'] = batchsize
comlen = len(seqdata['ctest'][list(seqdata['ctest'].keys())[0]])
config, _ = create_model(modeltype, config)
print("MODEL LOADED")
model = keras.models.load_model(modelfile, custom_objects={"tf":tf, "keras":keras,'AlexGraphLayer':GCNLayer})
node_data = seqdata['stest_nodes']
edgedata = seqdata['stest_edges']
config['batch_maker'] = 'graph_multi_1'
print(model.summary())
# set up prediction string and output file
comstart = np.zeros(comlen)
stk = comstok.w2i['<s>']
comstart[0] = stk
outfn = outdir+"/predictions/predict-{}.txt".format(modeltype)
outf = open(outfn, 'w')
print("writing to file: " + outfn)
batch_sets = [allfids[i:i+batchsize] for i in range(0, len(allfids), batchsize)]
index = 0
final_dict = collections.defaultdict(dict)
for c, fid_set in enumerate(batch_sets):
index += 1
if index == 20:
break
st = timer()
for fid in fid_set:
seqdata['ctest'][fid] = comstart #np.asarray([stk])
print("fid", fid_set)
print("c", c)
bg = batch_gen(seqdata, 'test', config, nodedata=node_data, edgedata=edgedata)
batch = bg.make_batch(fid_set)
batch_results, output_dict, dict1 = gen_pred(fid_set, model, batch, comstok, smltok, comlen, batchsize, config, strat='greedy')
final_dict[fid_set[0]]["layer_output"] = output_dict
final_dict[fid_set[0]]["input"] = dict1
for key, val in batch_results.items():
print("summary", val)
outf.write("{}\t{}\n".format(key, val))
end = timer ()
print("{} processed, {} per second this batch".format((c+1)*batchsize, int(batchsize/(end-st))), end='\r')
# break
with open('ICPC2020_GNN/modelout/predictions/final_dict.json', 'w') as fp:
# pickle.dump(dict1, fp)
fp.write(json.dumps(final_dict, cls=NDArrayEncoder))
outf.close()