-
Notifications
You must be signed in to change notification settings - Fork 20
/
run_bertrl.py
320 lines (270 loc) · 12.4 KB
/
run_bertrl.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
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
""" Finetuning BERTRL for scoring.
Adapted from huggingface transformers sequence classification on GLUE"""
import dataclasses
import logging
import os
import sys
from dataclasses import dataclass, field
from typing import Callable, Dict, Optional
import numpy as np
from transformers import AutoConfig, AutoModelForSequenceClassification, AutoTokenizer, EvalPrediction, GlueDataset
from transformers import GlueDataTrainingArguments as DataTrainingArguments
from transformers import (
HfArgumentParser,
Trainer,
TrainingArguments,
glue_compute_metrics,
glue_output_modes,
glue_tasks_num_labels,
set_seed,
)
import torch
from sklearn.metrics import average_precision_score
from collections import defaultdict, Counter
import itertools
from model import BertForKBCSequenceClassification
logger = logging.getLogger(__name__)
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
"""
model_name_or_path: str = field(
metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"}
)
config_name: Optional[str] = field(
default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"}
)
tokenizer_name: Optional[str] = field(
default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"}
)
cache_dir: Optional[str] = field(
default=None, metadata={"help": "Where do you want to store the pretrained models downloaded from s3"}
)
def main():
# See all possible arguments in src/transformers/training_args.py
# or by passing the --help flag to this script.
# We now keep distinct sets of args, for a cleaner separation of concerns.
parser = HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments))
if len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
# If we pass only one argument to the script and it's the path to a json file,
# let's parse it to get our arguments.
model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1]))
else:
model_args, data_args, training_args = parser.parse_args_into_dataclasses()
if (
os.path.exists(training_args.output_dir)
and os.listdir(training_args.output_dir)
and training_args.do_train
and not training_args.overwrite_output_dir
):
raise ValueError(
f"Output directory ({training_args.output_dir}) already exists and is not empty. Use --overwrite_output_dir to overcome."
)
# Setup logging
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO if training_args.local_rank in [-1, 0] else logging.WARN,
)
logger.warning(
"Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
training_args.local_rank,
training_args.device,
training_args.n_gpu,
bool(training_args.local_rank != -1),
training_args.fp16,
)
logger.info("Training/evaluation parameters %s", training_args)
# Set seed
set_seed(training_args.seed)
try:
num_labels = glue_tasks_num_labels[data_args.task_name]
output_mode = glue_output_modes[data_args.task_name]
except KeyError:
raise ValueError("Task not found: %s" % (data_args.task_name))
# Load pretrained model and tokenizer
#
# Distributed training:
# The .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
config = AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path,
num_labels=num_labels,
finetuning_task=data_args.task_name,
cache_dir=model_args.cache_dir,
)
tokenizer = AutoTokenizer.from_pretrained(
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path,
cache_dir=model_args.cache_dir,
use_fast=True,
)
# model = AutoModelForSequenceClassification.from_pretrained(
model = BertForKBCSequenceClassification.from_pretrained(
model_args.model_name_or_path,
from_tf=bool(".ckpt" in model_args.model_name_or_path),
config=config,
cache_dir=model_args.cache_dir,
)
# Get datasets
train_dataset = (
GlueDataset(data_args, tokenizer=tokenizer, cache_dir=model_args.cache_dir) if training_args.do_train else None
)
eval_dataset = (
GlueDataset(data_args, tokenizer=tokenizer, mode="dev", cache_dir=model_args.cache_dir)
if training_args.do_eval
else None
)
test_dataset = (
GlueDataset(data_args, tokenizer=tokenizer, mode="test", cache_dir=model_args.cache_dir)
if training_args.do_predict
else None
)
# evaluating BERTRL predictions in Hits and MRR.
def build_compute_metrics_fn(task_name: str) -> Callable[[EvalPrediction], Dict]:
def compute_ranking_metrics_fn(p: EvalPrediction):
t_res = torch.softmax(torch.from_numpy(p.predictions), 1)
eid2pos = {}
eid2neg = defaultdict(list)
bertrl_data_file = f'{data_args.data_dir}/dev.tsv' # bertrl data file (linearized)
examples = [l.strip().split('\t') for l in open(bertrl_data_file)][1:]
# another set of evaluation
eid2pos = defaultdict(list)
eid2neg = defaultdict(list)
eid2neg_lv2 = {}
for i, example in enumerate(examples):
label = int(example[0])
eid = example[1].split('-')[2] # test-neg-2-24
if label == 1:
eid2pos[eid].append(i)
else:
e_negid = example[1].split('-')[3] # test-neg-2-24
eid2neg[eid].append(i)
if eid not in eid2neg_lv2:
eid2neg_lv2[eid] = defaultdict(list)
eid2neg_lv2[eid][e_negid].append(i)
hit1 = 0
# hits = {1:0, 2:0, 3:0, 4:0, 5:0, 10:0, 20:0, 30:0, 40:0, 50:0}
hits = Counter()
ranks = []
for eid, pos_is in eid2pos.items():
pos_scores = t_res[pos_is, 1]
neg_is = eid2neg[eid]
geq_j = 1 # rank
if not neg_is: # no negative, only positive
hit1 += 1
else:
neg_scores_of_eid = []
neg_scores_lists = []
for neg_eid in eid2neg_lv2[eid]:
neg_is_of_eid = eid2neg_lv2[eid][neg_eid]
neg_scores_ = t_res[neg_is_of_eid, 1] # previously a bug here, previously use neg_is which lowers the hit 2+ performance
neg_max_score_ = torch.max(neg_scores_).item()
neg_scores_of_eid.append(neg_max_score_)
neg_scores_lists.append(neg_scores_.sort(0, descending=True).values.tolist())
_scores_pos = pos_scores.sort(0, descending=True).values.tolist()
for _scores in neg_scores_lists:
for s1, s2 in itertools.zip_longest(_scores_pos, _scores, fillvalue=100): # fill 100 as a default value
if s1 < s2 or s1 == 100:
geq_j += 1
break
elif s1 == s2:
continue
else:
break
for hit in [1, 2, 3, 4, 5, 10, 20, 30, 40, 50]:
if geq_j <= hit:
hits[hit] += 1
ranks.append(geq_j)
hits = {f'hit@{k}':hits[k] for k in sorted(hits)}
mrr = np.mean(1 / np.array(ranks))
hits['filtered_mrr'] = mrr
return hits
def compute_metrics_fn(p: EvalPrediction):
# print(compute_ranking_metrics_fn(p))
preds = p.predictions[0] if isinstance(p.predictions, tuple) else p.predictions
if output_mode == "classification":
preds = np.argmax(preds, axis=1)
else: # regression
preds = np.squeeze(preds)
glue_metrics = glue_compute_metrics(task_name, preds, p.label_ids)
glue_metrics.update(compute_ranking_metrics_fn(p))
return glue_metrics
# return compute_ranking_metrics_fn
return compute_metrics_fn
# Initialize our Trainer
trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_dataset,
eval_dataset=eval_dataset,
compute_metrics=build_compute_metrics_fn(data_args.task_name),
)
# Training
if training_args.do_train:
trainer.train(
model_path=model_args.model_name_or_path if os.path.isdir(model_args.model_name_or_path) else None
)
trainer.save_model()
# For convenience, we also re-save the tokenizer to the same directory,
# so that you can share your model easily on huggingface.co/models =)
if trainer.is_world_master():
tokenizer.save_pretrained(training_args.output_dir)
# customized early stopping
# Evaluation
eval_results = {}
if training_args.do_eval:
logger.info("*** Evaluate ***")
# Loop to handle MNLI double evaluation (matched, mis-matched)
eval_datasets = [eval_dataset]
if data_args.task_name == "mnli":
mnli_mm_data_args = dataclasses.replace(data_args, task_name="mnli-mm")
eval_datasets.append(
GlueDataset(mnli_mm_data_args, tokenizer=tokenizer, mode="dev", cache_dir=model_args.cache_dir)
)
for eval_dataset in eval_datasets:
trainer.compute_metrics = build_compute_metrics_fn(eval_dataset.args.task_name)
eval_result = trainer.evaluate(eval_dataset=eval_dataset)
# ranking evaluations
# preds = trainer.predict(test_dataset=eval_dataset).predictions
output_eval_file = os.path.join(
training_args.output_dir, f"eval_results_{eval_dataset.args.task_name}.txt"
)
if trainer.is_world_master():
with open(output_eval_file, "w") as writer:
logger.info("***** Eval results {} *****".format(eval_dataset.args.task_name))
for key, value in eval_result.items():
logger.info(" %s = %s", key, value)
writer.write("%s = %s\n" % (key, value))
eval_results.update(eval_result)
if training_args.do_predict:
logging.info("*** Test ***")
test_datasets = [test_dataset]
if data_args.task_name == "mnli":
mnli_mm_data_args = dataclasses.replace(data_args, task_name="mnli-mm")
test_datasets.append(
GlueDataset(mnli_mm_data_args, tokenizer=tokenizer, mode="test", cache_dir=model_args.cache_dir)
)
for test_dataset in test_datasets:
predictions = trainer.predict(test_dataset=test_dataset).predictions
if output_mode == "classification":
# write scores of each example into files
with open(f'{training_args.output_dir}/test_results_prediction_scores.npy', 'wb') as fout:
np.save(fout, predictions)
predictions = np.argmax(predictions, axis=1)
output_test_file = os.path.join(
training_args.output_dir, f"test_results_{test_dataset.args.task_name}.txt"
)
if trainer.is_world_master():
with open(output_test_file, "w") as writer:
logger.info("***** Test results {} *****".format(test_dataset.args.task_name))
writer.write("index\tprediction\n")
for index, item in enumerate(predictions):
if output_mode == "regression":
writer.write("%d\t%3.3f\n" % (index, item))
else:
item = test_dataset.get_labels()[item]
writer.write("%d\t%s\n" % (index, item))
return eval_results
if __name__ == "__main__":
main()