File size: 6,588 Bytes
67d6b31
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from torch.utils.data import DataLoader
from sentence_transformers import LoggingHandler
from sentence_transformers.cross_encoder import CrossEncoder
from sentence_transformers.cross_encoder.evaluation import CEBinaryClassificationEvaluator
from sentence_transformers import InputExample
import logging
from datetime import datetime
import gzip
import sys
import numpy as np
import os
from shutil import copyfile
import csv
import tqdm

#### Just some code to print debug information to stdout
logging.basicConfig(format='%(asctime)s - %(message)s',
                    datefmt='%Y-%m-%d %H:%M:%S',
                    level=logging.INFO,
                    handlers=[LoggingHandler()])
#### /print debug information to stdout


#Define our Cross-Encoder
model_name = sys.argv[1] #'google/electra-small-discriminator'
train_batch_size = 32
num_epochs = 1
model_save_path = 'output/training_ms-marco_cross-encoder-'+model_name.replace("/", "-")+'-'+datetime.now().strftime("%Y-%m-%d_%H-%M-%S")

#We set num_labels=1, which predicts a continous score between 0 and 1
model = CrossEncoder(model_name, num_labels=1, max_length=512)


# Write self to path
os.makedirs(model_save_path, exist_ok=True)

train_script_path = os.path.join(model_save_path, 'train_script.py')
copyfile(__file__, train_script_path)
with open(train_script_path, 'a') as fOut:
    fOut.write("\n\n# Script was called via:\n#python " + " ".join(sys.argv))


corpus = {}
queries = {}

#### Read train file
with gzip.open('../data/collection.tsv.gz', 'rt') as fIn:
    for line in fIn:
        pid, passage = line.strip().split("\t")
        corpus[pid] = passage

with open('../data/queries.train.tsv', 'r') as fIn:
    for line in fIn:
        qid, query = line.strip().split("\t")
        queries[qid] = query



pos_neg_ration = (4+1)
cnt = 0
train_samples = []
dev_samples = {}

num_dev_queries = 125
num_max_dev_negatives = 200

with gzip.open('../data/qidpidtriples.rnd-shuf.train-eval.tsv.gz', 'rt') as fIn:
    for line in fIn:
        qid, pos_id, neg_id = line.strip().split()

        if qid not in dev_samples and len(dev_samples) < num_dev_queries:
            dev_samples[qid] = {'query': queries[qid], 'positive': set(), 'negative': set()}

        if qid in dev_samples:
            dev_samples[qid]['positive'].add(corpus[pos_id])

            if len(dev_samples[qid]['negative']) < num_max_dev_negatives:
                dev_samples[qid]['negative'].add(corpus[neg_id])

with gzip.open('../data/qidpidtriples.rnd-shuf.train.tsv.gz', 'rt') as fIn:
    for line in tqdm.tqdm(fIn, unit_scale=True):
        cnt += 1
        qid, pos_id, neg_id = line.strip().split()
        query = queries[qid]
        if (cnt % pos_neg_ration) == 0:
            passage = corpus[pos_id]
            label = 1
        else:
            passage = corpus[neg_id]
            label = 0

        train_samples.append(InputExample(texts=[query, passage], label=label))

        if len(train_samples) >= 2e7:
            break



train_dataloader = DataLoader(train_samples, shuffle=True, batch_size=train_batch_size)

# We add an evaluator, which evaluates the performance during training

class CERerankingEvaluator:
    def __init__(self, samples, mrr_at_k: int = 10, name: str = ''):
        self.samples = samples
        self.name = name
        self.mrr_at_k = mrr_at_k

        if isinstance(self.samples, dict):
            self.samples = list(self.samples.values())

        self.csv_file = "CERerankingEvaluator" + ("_" + name if name else '') + "_results.csv"
        self.csv_headers = ["epoch", "steps", "MRR@{}".format(mrr_at_k)]

    def __call__(self, model, output_path: str = None, epoch: int = -1, steps: int = -1) -> float:
        if epoch != -1:
            if steps == -1:
                out_txt = " after epoch {}:".format(epoch)
            else:
                out_txt = " in epoch {} after {} steps:".format(epoch, steps)
        else:
            out_txt = ":"

        logging.info("CERerankingEvaluator: Evaluating the model on " + self.name + " dataset" + out_txt)

        all_mrr_scores = []
        num_queries = 0
        num_positives = []
        num_negatives = []
        for instance in self.samples:
            query = instance['query']
            positive = list(instance['positive'])
            negative = list(instance['negative'])
            docs = positive + negative
            is_relevant = [True]*len(positive) + [False]*len(negative)

            if len(positive) == 0 or len(negative) == 0:
                continue

            num_queries += 1
            num_positives.append(len(positive))
            num_negatives.append(len(negative))

            model_input = [[query, doc] for doc in docs]
            pred_scores = model.predict(model_input, convert_to_numpy=True, show_progress_bar=False)
            pred_scores_argsort = np.argsort(-pred_scores)  #Sort in decreasing order

            mrr_score = 0
            for rank, index in enumerate(pred_scores_argsort[0:self.mrr_at_k]):
                if is_relevant[index]:
                    mrr_score = 1 / (rank+1)

            all_mrr_scores.append(mrr_score)

        mean_mrr = np.mean(all_mrr_scores)
        logging.info("Queries: {} \t Positives: Min {:.1f}, Mean {:.1f}, Max {:.1f} \t Negatives: Min {:.1f}, Mean {:.1f}, Max {:.1f}".format(num_queries, np.min(num_positives), np.mean(num_positives), np.max(num_positives), np.min(num_negatives), np.mean(num_negatives), np.max(num_negatives)))
        logging.info("MRR@{}: {:.2f}".format(self.mrr_at_k, mean_mrr*100))

        if output_path is not None:
            csv_path = os.path.join(output_path, self.csv_file)
            output_file_exists = os.path.isfile(csv_path)
            with open(csv_path, mode="a" if output_file_exists else 'w', encoding="utf-8") as f:
                writer = csv.writer(f)
                if not output_file_exists:
                    writer.writerow(self.csv_headers)

                writer.writerow([epoch, steps, mean_mrr])

        return mean_mrr


evaluator = CERerankingEvaluator(dev_samples)

# Configure the training
warmup_steps = 5000
logging.info("Warmup-steps: {}".format(warmup_steps))


# Train the model
model.fit(train_dataloader=train_dataloader,
          evaluator=evaluator,
          epochs=num_epochs,
          evaluation_steps=5000,
          warmup_steps=warmup_steps,
          output_path=model_save_path,
          use_amp=True)

#Save latest model
model.save(model_save_path+'-latest')


# Script was called via:
#python train_cross-encoder.py nreimers/TinyBERT_L-6_H-768_v2