-
Notifications
You must be signed in to change notification settings - Fork 2
/
validate_integrated.py
185 lines (151 loc) · 7.15 KB
/
validate_integrated.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
import os
import sys
import random
import argparse
from datetime import datetime
import numpy as np
# PyTorch
import torch
import torch.nn as nn
import torch.optim as optim
# HuggingFace
from transformers import BartTokenizer
from datasets import load_dataset
# This project
from models.efficient_decoder_expC_integrated import BartEfficientDecoder
from utils import get_boundary_matrix, shifted_target_left
from utils import parse_config
def validate(args):
config_path = args['config_path']
model_path = args['load']
cache_dir = args['cache_dir']
start_id = args['start_id']
end_id = args['end_id']
# Load Config
config = parse_config("config", config_path)
# uses GPU in training or not
if torch.cuda.is_available() and args['use_gpu']: torch_device = 'cuda'
else: torch_device = 'cpu'
num_heads = config['num_heads']
num_layers = config['num_layers']
eos_id = config['eos_id']
batch_size = config['batch_size']
temperature = config['temperature']
eps = config['eps']
task = config['task'] # CNNDM | XSUM
max_target_len = config['max_target_len']
num_encoder_sent_nn_layers = config['num_encoder_sent_nn_layers']
training_ref = config['training_ref'] # exact, approx, mix
num_sentences_kept = config['r_train']
lambda1 = config['lambda1']
bart_tokenizer = BartTokenizer.from_pretrained(config['bart_tokenizer'])
bart = BartEfficientDecoder.from_pretrained(config['bart_weights'])
bart.sent_nn_init(num_encoder_sent_nn_layers)
bart.swap_crossattn_to_hier(num_layers)
if torch_device == 'cuda': state = torch.load(model_path)
else: state = torch.load(model_path, map_location=torch.device('cpu'))
model_state_dict = state['model']
bart.load_state_dict(model_state_dict)
print('loaded model from:', model_path)
if torch_device == 'cuda': bart.cuda()
del state, model_state_dict
torch.cuda.empty_cache()
bart_config = bart.config
bart.config.output_attentions = True
bart.model.encoder.output_attentions = True
bart.model.decoder.output_attentions = True
for i in range(num_layers):
bart.model.encoder.layers[i].output_attentions = True
bart.model.decoder.layers[i].output_attentions = True
if task == 'CNNDM':
valid_data = load_dataset('cnn_dailymail', '3.0.0', split='validation')
print("cnndm data loaded")
elif task == 'XSUM':
valid_data = load_dataset('xsum', split='validation')
print("xsum data loaded")
else:
raise ValueError("task not supported, only CNNDM | XSUM")
# Criterion
criterion = nn.CrossEntropyLoss(reduction='none') # This criterion combines nn.LogSoftmax() and nn.NLLLoss() in one single class.
criterion_kl = nn.KLDivLoss(reduction='mean')
assert batch_size == 1, "batch_size > 1 not supported"
ids = [_i for _i in range(start_id, end_id)]
if args['random_order']: random.shuffle(ids)
bart.eval()
for id in ids:
outpath = "{}/{}_vloss.txt".format(cache_dir, id)
exist = os.path.isfile(outpath)
if exist:
print("id {}: already exists".format(id))
continue
if task == 'CNNDM':
document = valid_data[i]['article']
summary = valid_data[i]['highlights']
elif task == 'XSUM':
document = valid_data[i]['document']
summary = valid_data[i]['summary']
try:
# some training data instances are corrupted
batch_encoded_inputs = bart_tokenizer.batch_encode_plus([document], return_tensors='pt',
add_special_tokens=True, max_length=bart.config.max_position_embeddings, pad_to_max_length=False)
input_ids = batch_encoded_inputs['input_ids'].to(torch_device)
batch_encoded_target = bart_tokenizer.batch_encode_plus([summary], return_tensors='pt',
add_special_tokens=True, max_length=max_target_len, pad_to_max_length=False)
target_ids = batch_encoded_target['input_ids'].to(torch_device)
except IndexError:
print("[IndexError]: id = {}".format(i))
continue
shifted_target_ids = shifted_target_left(target_ids)
# find boundaries
_input_ids = input_ids[0].cpu().numpy()
# boundary['boundary'].shape => [num_head, num_words, num_sent]
boundary = get_boundary_matrix(_input_ids, num_heads*batch_size, eos_id, torch_device)
if training_ref == 'exact':
boundary['use_exact_scores'] = True
elif training_ref == 'approx':
boundary['use_exact_scores'] = False
elif training_ref == 'mix':
raise Exception("Change training_ref to exact or approx for validation!!")
boundary['num_sentences_kept'] = num_sentences_kept
bart_outputs = bart(input_ids=input_ids, decoder_input_ids=target_ids, boundary=boundary)
lm_logits = bart_outputs[0]
sent_attns = bart_outputs[1]
num_sentences = len(boundary['sentid2wordid'])
loss_kl = 0
for l in range(num_layers):
exact_attn = sent_attns[l]['exact_sent_attn_weights'].squeeze(0) # # (num_heads, tgt_len, num_sentences)
exact_attn_hot = torch.softmax(torch.log(exact_attn + eps) / temperature, dim=-1)
rnn_attn = sent_attns[l]['rnn_sent_attn_weights'].squeeze(0) # # (num_heads, tgt_len, num_sentences)
loss_kl += criterion_kl(torch.log(rnn_attn + eps).view(-1, num_sentences), exact_attn_hot.view(-1, num_sentences))
if torch.isinf(loss_kl): # this works for both +ve and -ve inf
print("loss_kl is inf: id = {}".format(id))
raise("INF ERROR:", id)
if torch.isnan(loss_kl):
print("loss_kl is NaN: id = {}".format(id))
raise("NAN ERROR:", id)
loss_kl = loss_kl.item()
# Main Loss (cross entropy)
loss = criterion(lm_logits.view(-1, bart_config.vocab_size), shifted_target_ids.view(-1))
loss = loss.mean().item()
total_loss = loss + lambda1*loss_kl
with open(outpath, 'w') as f:
f.write("{:.8f}\n{:.8f}\n{:.8f}".format(total_loss, loss, loss_kl))
print("wrote:", outpath)
def get_decode_arguments(parser):
'''Arguments for decoding'''
parser.register("type", "bool", lambda v: v.lower() == "true")
# file paths
parser.add_argument('--load', type=str, required=True) # path to load model
parser.add_argument('--config_path',type=str, required=True) # path to load model
parser.add_argument('--cache_dir', type=str, required=True)
parser.add_argument('--start_id', type=int, required=True)
parser.add_argument('--end_id', type=int, required=True)
parser.add_argument('--random_order', type="bool", nargs="?", const=True, default=False)
parser.add_argument('--use_gpu', type="bool", nargs="?", const=True, default=False)
return parser
if __name__ == "__main__":
# get configurations from the terminal
parser = argparse.ArgumentParser()
parser = get_decode_arguments(parser)
args = vars(parser.parse_args())
validate(args)