-
Notifications
You must be signed in to change notification settings - Fork 1
/
eval_baseline_ppl.py
339 lines (284 loc) · 13.9 KB
/
eval_baseline_ppl.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
321
322
323
324
325
326
327
328
329
330
331
332
import logging
import logzero
import argparse
import os
import string
import time
import random
import numpy as np
import re
from statistics import mean
from tqdm import tqdm
from collections import Counter
import csv
from pathlib import Path
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.utils.data import DataLoader
from transformers import AutoTokenizer
from transformers import AutoModelForMaskedLM
from utils import load
from datasets import CreoleJsonDataset, CreoleDatasetWILDS
from wilds.common.grouper import CombinatorialGrouper
from wilds.common.data_loaders import GroupSampler
from wilds.common.utils import get_counts
from wilds.common.metrics.loss import ElementwiseLoss, MultiTaskLoss
from wilds.common.metrics.all_metrics import Accuracy, MultiTaskAccuracy, MSE
from algorithms.groupDRO import GroupDRO
def parse_args():
parser = argparse.ArgumentParser()
# Data
parser.add_argument("--file_path", type=str, default="",
help="Path to the data you are trying to finetune on or evaluate from")
parser.add_argument("--dictionary_path", type=str, default="",
help="Path to the creole specific dictionary")
parser.add_argument("--creole", type=str, default="", choices=["singlish", "haitian", "naija"])
parser.add_argument("--experiment", type=str, default="baseline", choices=["pretrained", "baseline", "dro"])
parser.add_argument("--group_strategy", type=str, default="collect",
choices=["collect", "cluster", "percent", "random", "one", "language"])
# Model
parser.add_argument("--tokenizer", type=str, default='bert-base-uncased',
help="Pretrained BERT: bert-base-uncased, bert-base-multilingual-cased, xlm-roberta-base, etc.")
parser.add_argument("--from_pretrained", type=str, default='bert-base-uncased',
help="Pretrained BERT: bert-base-uncased, bert-base-multilingual-cased, xlm-roberta-base, etc.,"
"Or full path to our pretrained model.")
parser.add_argument("--base_lang", type=str, default="en",
help="Base language of the Creole. en or fr")
# Logging
parser.add_argument("--checkpoint_dir", type=str, default="")
# Eval
parser.add_argument("--seed", type=int, default=42)
parser.add_argument("--batch_size", type=int, default=1)
#DRO stuff
# Group
parser.add_argument("--algo_log_metric", type=str, default="mse")
parser.add_argument("--train_loader", type=str, default="standard", choices=['standard', 'group'])
parser.add_argument("--uniform_over_groups", default=True, action="store_true")
parser.add_argument("--n_groups_per_batch", type=int, default=1)
parser.add_argument("--no_group_logging", default=True, action="store_true")
parser.add_argument("--group_dro_step_size", type=float, default=0.01)
# Training
parser.add_argument("--device", type=str, default="cuda")
parser.add_argument("--loss_function", type=str, default="cross_entropy",
choices=["cross_entropy", "mse", "multitask_bcd"])
parser.add_argument("--num_epochs", type=int, default=1)
parser.add_argument("--learning_rate", type=float, default=2e-5,
help="Former default was 5e-5")
parser.add_argument("--optimizer", type=str, default="AdamW")
parser.add_argument("--weight_decay", type=float, default=0.01)
parser.add_argument("--adam_epsilon", type=float, default=1e-8)
parser.add_argument("--max_grad_norm", type=int, default=1.0)
parser.add_argument("--scheduler", type=str, default="linear_schedule_with_warmup")
parser.add_argument("--scheduler_metric_name", type=str, default="fuckoff")
parser.add_argument("--num_warmup_steps", type=int, default=0,
help="Default in run_glue.py")
return parser.parse_args()
def mask_dict_word(tokens, tokenizer, mlm_idxs):
output_label = [-100] * len(tokens)
for i in mlm_idxs:
output_label[i] = tokens[i].item()
tokens[i] = tokenizer.convert_tokens_to_ids(tokenizer.mask_token)
return tokens.unsqueeze(0), torch.LongTensor(output_label).unsqueeze(0), mlm_idxs
def mask_1word(tokens, tokenizer):
output_label = [-100] * len(tokens)
rnd_token_ix = random.choice(np.arange(1, torch.where(tokens == 102)[0][0].item()))
output_label[rnd_token_ix] = tokens[rnd_token_ix].item()
tokens[rnd_token_ix] = tokenizer.convert_tokens_to_ids(tokenizer.mask_token)
return tokens.unsqueeze(0), torch.LongTensor(output_label).unsqueeze(0), rnd_token_ix
def mask_allwords(tokens, tokenizer):
max_ix = torch.where(tokens == 102)[0][0].item()
batch_ids = torch.zeros(max_ix-1, tokens.size(0), dtype=torch.long)
output_labels = torch.zeros_like(batch_ids) - 100
mask_ix = tokenizer.convert_tokens_to_ids(tokenizer.mask_token)
for ix, tok_ix in enumerate(np.arange(1, max_ix)):
sent = tokens.clone()
out = tokens[ix+1].item()
sent[ix+1] = mask_ix
batch_ids[ix] = sent
output_labels[ix, ix+1] = out
return batch_ids, output_labels
def is_sublist(a, b, start):
if not a: return True, "fuck"
if not b: return False, "blah"
if b[:len(a)] == a:
return start, len(a)
else:
return is_sublist(a, b[1:], start+1)
#return b[:len(a)] == a or is_sublist(a, b[1:])
def get_model_at_epoch_evals(model, eval_dataset, experiment, args):
# PPL
print("Computing PLL...")
csv_columns = ["experiment", "i_example", "runningPPL"]
results_dir = "/science/image/nlp-datasets/creoles/pplresults"
Path(results_dir).mkdir(parents=True, exist_ok=True)
if "mixed" in experiment:
results_file = f"mixed_{args.creole}.csv"
elif "creoleonly" in experiment:
results_file = f"creoleonly_{args.creole}.csv"
else:
results_file = f"results_{args.creole}.csv"
full_path_to_results = os.path.join(results_dir, results_file)
ppls = []
for i in tqdm(range(len(eval_dataset))):
# Get tokenized sentence
sent = eval_dataset.__getitem__(i)
# Mask all tokens
tokens, output_labels = mask_allwords(sent, tokenizer)
tokens = tokens.to(device)
output_labels = output_labels.to(device)
with torch.no_grad():
result = model(tokens, token_type_ids=None, labels=output_labels)
lm_loss = F.cross_entropy(result.logits.view(-1, tokenizer.vocab_size), output_labels.view(-1), reduction="sum")
ppls.append(lm_loss.cpu().item())
row = [experiment, len(ppls), round(np.mean(ppls), 4)]
if not os.path.isfile(full_path_to_results):
with open(full_path_to_results, 'w') as file:
writer = csv.writer(file)
writer.writerow(csv_columns)
writer.writerow(row)
else: # append results
with open(full_path_to_results, 'a') as file:
writer = csv.writer(file)
writer.writerow(row)
mean_ppl = np.mean(ppls)
print(f"Mean PLL: {mean_ppl}")
row = [experiment, "FINAL", round(mean_ppl,4)]
with open(full_path_to_results, 'a') as file:
writer = csv.writer(file)
writer.writerow(row)
return round(mean_ppl,4)
def clean_checkpoint_name(filepath):
return filepath.split("conll")[-1]
if __name__ == "__main__":
args = parse_args()
# log.debug(args)
#log_level = logging.DEBUG if args.debug else logging.INFO
#logzero.loglevel(log_level)
#logzero.formatter(logzero.LogFormatter(datefmt="%Y-%m-%d %H:%M:%S"))
# load BERT tokenizer
print('Loading tokenizer...')
tokenizer = AutoTokenizer.from_pretrained(args.tokenizer, do_lower_case=('uncased' in args.from_pretrained))
random.seed(args.seed)
np.random.seed(args.seed)
torch.manual_seed(args.seed)
if torch.cuda.is_available():
torch.cuda.manual_seed_all(args.seed)
# If there's a GPU available...
if torch.cuda.is_available():
device = torch.device("cuda")
print('There are %d GPU(s) available.' % torch.cuda.device_count())
print('We will use the GPU:', torch.cuda.get_device_name(0))
# If not...
else:
print('No GPU available, using the CPU instead.')
device = torch.device("cpu")
# Load evaluation data
eval_dataset = CreoleJsonDataset(src_file=args.file_path, tokenizer=tokenizer, base_language=args.base_lang, creole_only=True)
"""
if args.creole == "singlish":
eval_dataset = SinglishUDDataset(src_dir=args.file_path, tokenizer=tokenizer)
creole_dictionary = creole_dict_reader(path=args.dictionary_path)
elif args.creole == "naija":
creole_dictionary = creole_dict_reader(path=args.dictionary_path)
if "SUD" in args.file_path:
eval_dataset = NaijaUDDataset(src_dir=args.file_path, tokenizer=tokenizer)
elif "masakhane" in args.file_path:
eval_dataset = NaijaMasakhaneNERDataset(src_dir=args.file_path, tokenizer=tokenizer)
else:
raise NotImplementedError
elif args.creole == "haitian":
eval_dataset = HaitianEvalDatasets(src_dir=args.file_path, tokenizer=tokenizer)
creole_dictionary = haitian_dict_reader(path=args.dictionary_path)
else:
print(f"please specify the argument --creole= from ['singlish', 'haitian', 'naija']")
print(f"other creoles have not been implemented")
raise NotImplementedError
"""
if args.experiment == "baseline":
model_dirs = os.listdir(args.checkpoint_dir)
model_dirs = sorted([d for d in model_dirs if d.isnumeric() and d in ["100000"]])
for epoch in model_dirs:
full_path = os.path.join(args.checkpoint_dir, epoch)
model = AutoModelForMaskedLM.from_pretrained(full_path)
model.to(device)
model.eval()
final_ppl = get_model_at_epoch_evals(model, eval_dataset, full_path, args)
del model
if args.experiment == "dro":
args.optimizer_kwargs = {'eps': 1e-8}
args.scheduler_kwargs = {'num_warmup_steps': 0}
vocab_size = tokenizer.vocab_size
train_dataset = CreoleDatasetWILDS(eval_dataset, tokenizer, group_strategy="one",
group_file="",
creole=args.creole) # this one has (x, y, metadata)
train_grouper = CombinatorialGrouper(dataset=train_dataset, groupby_fields=train_dataset.metadata_fields)
group_ids = train_grouper.metadata_to_group(train_dataset.metadata_array)
batch_sampler = GroupSampler(
group_ids=group_ids,
batch_size=args.batch_size,
n_groups_per_batch=train_grouper.n_groups,
uniform_over_groups=False, # was True
distinct_groups=False) # was True
torch.set_printoptions(threshold=100)
print(f"group_ids: {group_ids} | num groups: {train_grouper.n_groups}")
print(f"size of groups: {group_ids.size()}")
print(Counter(group_ids.tolist()))
train_loader = DataLoader(train_dataset, shuffle=False, sampler=None, batch_sampler=batch_sampler,
drop_last=False)
print(f"metadata_array: {train_dataset.metadata_array}")
train_g = train_grouper.metadata_to_group(train_dataset.metadata_array)
is_group_in_train = get_counts(train_g, train_grouper.n_groups) > 0
print(f"is_group_in_train: {is_group_in_train}")
# init DRO algorithm
base_model = AutoModelForMaskedLM.from_pretrained(args.from_pretrained).to(device)
#print(f"base model: {base_model}")
# options for losses and metric
losses = {
'cross_entropy': ElementwiseLoss(loss_fn=nn.CrossEntropyLoss(reduction='none', ignore_index=-100)),
'mse': MSE(name='loss'),
'multitask_bce': MultiTaskLoss(loss_fn=nn.BCEWithLogitsLoss(reduction='none')),
}
algo_log_metrics = {
'accuracy': Accuracy(),
'mse': MSE(),
'multitask_accuracy': MultiTaskAccuracy(),
# 'f1': F1(average='macro'),
None: None,
}
algorithm = GroupDRO(
config=args,
model=base_model,
d_out=train_dataset.y_size,
grouper=train_grouper,
loss=losses[args.loss_function], # cross_entropy
metric=None, # MSE(), #
n_train_steps=len(train_loader) * args.num_epochs,
is_group_in_train=is_group_in_train)
#dirLUT = {'bert-base-uncased': 'bert', 'bert-base-multilingual-cased': 'mbert', 'xlm-roberta-base': 'xlmr'}
#path_to_checkpoint = f"{args.checkpoint_dir}/{dirLUT[args.tokenizer]}/{args.creole}"
all_the_models = os.listdir(args.checkpoint_dir)
selected_models = sorted([m for m in all_the_models if args.group_strategy in m])
look_up_models = []
for model in selected_models:
if any(e in model for e in ["100000"]):
look_up_models.append(model)
print(f"look up models: {look_up_models}")
for cached_model in look_up_models:
epoch_number = 100000 #cached_model[-5]
full_path_to_model = os.path.join(args.checkpoint_dir, cached_model)
print(f"check path: {full_path_to_model}")
algorithm, epoch = load(algorithm=algorithm, path=full_path_to_model, device=device)
print(f"epoch: {epoch}")
algorithm.to(device)
algorithm.eval()
#exit(333)
model = algorithm.model
final_ppl = get_model_at_epoch_evals(model, eval_dataset, full_path_to_model, args)
del model
if args.experiment == "pretrained":
model = AutoModelForMaskedLM.from_pretrained(args.from_pretrained)
model.to(device)
model.eval()
final_ppl = get_model_at_epoch_evals(model, eval_dataset, args.from_pretrained, args)
del model