-
Notifications
You must be signed in to change notification settings - Fork 2
/
main_qaoe_tsv_mlm_head.py
209 lines (178 loc) · 9.46 KB
/
main_qaoe_tsv_mlm_head.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
from utils.lib import *
from dataset import get_tsv_dls
from main_qaoe_tsv import Dataset_QAOE_TSV
from main_qaoe_lsmdc_fib import Agent_QAOE_LSMDC
from utils.args import get_args
from utils.logger import LOGGER, add_log_to_file
from utils.dist import NoOp, is_main_process, all_gather, get_rank, get_world_size, iter_tqdm
from main_qaoe_lsmdc_fib import VIOLET_QAOE_LSMDC as VIOLET_QAOE_MLM_Head
from tqdm import tqdm
class Dataset_QAOE_MLM_Head(Dataset_QAOE_TSV):
def __init__(self, args, img_tsv_path, txt, id2lineidx, split, tokzr=None):
super().__init__(args, img_tsv_path, txt, id2lineidx, split, tokzr=tokzr)
total_examples = len(self.txt)
invalid_examples = 0
for item in self.txt:
ans = item['answer_text']
ans_id = self.tokzr.convert_tokens_to_ids([ans])[0]
if ans_id==self.unk_token_id: invalid_examples += 1
LOGGER.info(f"Split {split}, Invalid examples: {invalid_examples} "
f"/ Total examples: {total_examples}, "
f"upper-bound: {(1 - invalid_examples/total_examples)*100:.2f}%")
def append_mask(self, tokens, padding_len):
tokens = [self.tokzr.cls_token]+tokens+self.tokzr.tokenize(f"answer: ")+[self.tokzr.mask_token, self.tokzr.sep_token]+[self.tokzr.pad_token]*(padding_len)
return tokens
def prepend_mask(self, tokens, padding_len):
tokens = [self.tokzr.mask_token]+[self.tokzr.cls_token]+tokens+[self.tokzr.sep_token]+[self.tokzr.pad_token]*(padding_len)
return tokens
def replace_cls(self, tokens, padding_len):
tokens = [self.tokzr.mask_token]+tokens+[self.tokzr.sep_token]+[self.tokzr.pad_token]*(padding_len)
return tokens
def insert_mask(self, tokens, padding_len):
tokens = [self.tokzr.cls_token]+tokens+[self.tokzr.sep_token]+[self.tokzr.pad_token]*(padding_len)
if len(tokens)<10: tokens += [self.tokzr.mask_token]
else: tokens = tokens[:10]+[self.tokzr.mask_token]+tokens[10:]
return tokens
def str2txt(self, s):
tokens = self.tokzr.tokenize(s)
tokens = tokens[:self.args.size_txt-1]
padding_len = self.args.size_txt-len(tokens)
if self.args.mask_pos=="append": tokens = self.append_mask(tokens, padding_len)
elif self.args.mask_pos=="prepend": tokens = self.prepend_mask(tokens, padding_len)
elif self.args.mask_pos=="insert": tokens = self.insert_mask(tokens, padding_len)
elif self.args.mask_pos=="replace": tokens = self.replace_cls(tokens, padding_len)
txt = self.tokzr.convert_tokens_to_ids(tokens)
mask = [1 if w!=self.pad_token_id else 0 for w in txt]
mask = T.LongTensor(mask)
txt = T.LongTensor(txt)
return txt, mask
@property
def prompt_text(self):
return "answer the question about the video."
def __getitem__(self, idx):
item = self.txt[idx]
video_id = item['video']
if video_id in self.id2lineidx:
lineidx = self.id2lineidx[video_id]
b = self.seek_img_tsv(lineidx)[2:]
img = self.get_img_or_video(b)
else:
print(f"video missing: {video_id}")
img = T.zeros((self.args.size_frame, 3, self.args.size_img, self.args.size_img))
q = item['question']
txt, mask = self.str2txt(q)
if self.args.size_vocab>0: ans_id = item['answer']
else:
ans = item['answer_text']
ans_id = self.tokzr.convert_tokens_to_ids([ans])[0]
if ans_id==self.unk_token_id: ans_id = -1
if video_id not in self.id2lineidx:
print(f"video {video_id} not found")
ans_id = -1
mask_ans = T.ones(txt.shape).long()*-1
mask_ans[txt==self.mask_token_id] = ans_id
return img, txt, mask, mask_ans
def collate_batch(self, inputs):
img, txt, mask, mask_ans = map(list, unzip(inputs))
all_imgs = T.stack(img, dim=0)
all_mask_ans = T.stack(mask_ans, dim=0)
all_txts = T.stack(txt, dim=0)
all_masks = T.stack(mask, dim=0)
batch = {"img": all_imgs, "txt": all_txts, "mask": all_masks, "mask_ans": all_mask_ans}
return batch
class Agent_QAOE_MLM_Head(Agent_QAOE_LSMDC):
def __init__(self, args, model):
super().__init__(args, model)
if args.freeze_violet: self.model.freeze()
def go_dl(self, ep, dl, is_train):
if is_train: self.model.train()
else: self.model.eval()
ret = defaultdict(list)
idx = 0
for idx, batch in iter_tqdm(enumerate(dl)):
if is_train: self.global_step += 1
if (idx%self.args.logging_steps)==0 and is_train: LOGGER.info(self.log_memory(ep, idx+1))
if self.args.enable_prompt: batch["prompt"] = dl.dataset.get_prompt()
elif self.args.enable_task_token: batch["task_name"] = "oe"
batch = self.prepare_batch(batch)
r = self.step(batch, is_train)
if is_train: self.log_dict_to_wandb({f"train_{k}": l for k, l in r.items()})
ret = {k: ret[k]+l if isinstance(l, list) else ret[k]+[l] for k, l in r.items()}
if (idx%self.args.logging_steps)!=0 and is_train: LOGGER.info(self.log_memory(ep, idx+1))
gathered_ret = defaultdict(list)
for ret_per_rank in all_gather(ret):
for k in ret_per_rank: gathered_ret[k].extend(ret_per_rank[k])
ret_all = {k: float(np.average(gathered_ret[k])) for k in ret}
return ret_all
if __name__=='__main__':
args = get_args()
args.size_vocab = -1
tokzr = transformers.AutoTokenizer.from_pretrained(args.tokenizer)
dl_tr, dl_vl, dl_ts = get_tsv_dls(args, Dataset_QAOE_MLM_Head, tokzr=tokzr)
print(len(dl_tr), len(dl_vl), len(dl_ts))
if args.size_epoch==0: args.max_iter = 1
else: args.max_iter = len(dl_tr)*args.size_epoch
args.actual_size_test = len(dl_ts.dataset)
model = VIOLET_QAOE_MLM_Head(args, tokzr=tokzr)
model.load_ckpt(args.path_ckpt)
if args.reinit_head: model.reinit_head()
model.cuda()
if args.distributed: LOGGER.info(f"n_gpu: {args.num_gpus}, rank: {get_rank()},"
f" world_size: {get_world_size()}")
args.path_output = '%s/_%s_%s'%(args.path_output, args.task, datetime.now().strftime('%Y%m%d%H%M%S'))
agent = Agent_QAOE_MLM_Head(args, model)
if args.distributed: agent.prepare_dist_model()
agent.save_training_meta()
if is_main_process(): add_log_to_file('%s/stdout.txt' % (args.path_output))
else: LOGGER = NoOp()
LOGGER.info("Saved training meta infomation...")
if os.path.exists(args.path_ckpt):
LOGGER.info("Zero-shot Evaluation")
if len(dl_vl):
ac_vl = agent.go_dl(0, dl_vl, False)
LOGGER.info(f'ZS (val): {ac_vl["ac_1"]*100:.2f}, {ac_vl["ac_5"]*100:.2f}')
print('ZS (val):', ac_vl["ac_1"], ac_vl["ac_5"])
if len(dl_ts):
ac_ts = agent.go_dl(0, dl_ts, False)
LOGGER.info(f'ZS (test): {ac_ts["ac_1"]*100:.2f}, {ac_ts["ac_5"]*100:.2f}')
print('ZS (test):', ac_ts["ac_1"], ac_ts["ac_5"])
if (hasattr(args, "size_test") and args.size_test!=args.actual_size_test):
adjusted_ac_ts_1 = ac_ts['ac_1']*args.actual_size_test/args.size_test*100
adjusted_ac_ts_5 = ac_ts['ac_5']*args.actual_size_test/args.size_test*100
LOGGER.info(f'ZS (test, adjusted): {adjusted_ac_ts_1:.2f}'
f', {adjusted_ac_ts_5:.2f}')
print('ZS (test, adjusted):', adjusted_ac_ts_1, adjusted_ac_ts_5)
else: LOGGER.info("No pre-trained weight, skip zero-shot Evaluation")
if args.size_epoch:
agent.setup_wandb()
LOGGER.info("Start training....")
for e in iter_tqdm(range(args.size_epoch)):
ls_tr = agent.go_dl(e+1, dl_tr, True)
LOGGER.info(f'Ep {e}, Loss (train): {ls_tr["ls"]*100:.4e}')
if len(dl_vl):
ac_vl = agent.go_dl(e+1, dl_vl, False)
for k in ac_vl:
agent.log[f'{k}_vl'].append(ac_vl[k])
agent.log_dict_to_wandb({"{k}_vl": ac_vl[k]})
LOGGER.info(f'Ep {e}, Acc (val): {ac_vl["ac_1"]*100:.2f}, '
f'{ac_vl["ac_5"]*100:.2f}')
if len(dl_ts):
ac_ts = agent.go_dl(e+1, dl_ts, False)
LOGGER.info(f'Ep {e}, Acc (test): {ac_ts["ac_1"]*100:.2f}, '
f'{ac_ts["ac_5"]*100:.2f}')
if (hasattr(args, "size_test") and args.size_test!=args.actual_size_test):
adjusted_ac_ts_1 = ac_ts['ac_1']*args.actual_size_test/args.size_test
adjusted_ac_ts_5 = ac_ts['ac_5']*args.actual_size_test/args.size_test
agent.log['ac_1_ts'].append(adjusted_ac_ts_1)
agent.log['ac_5_ts'].append(adjusted_ac_ts_5)
LOGGER.info(f'Ep {e}, Acc (test, adjusted): {adjusted_ac_ts_1*100:.2f}'
f', {adjusted_ac_ts_5*100:.2f}')
else:
for k in ac_ts:
agent.log_dict_to_wandb({"{k}_ts": ac_ts[k]})
agent.log[f'{k}_ts'].append(ac_ts[k])
agent.save_model(e+1)
best_vl, best_ts = agent.best_epoch()
LOGGER.info(f'Best val @ ep {best_vl[0]+1}, {best_vl[1]*100:.2f}')
LOGGER.info(f'Best test @ ep {best_ts[0]+1}, {best_ts[1]*100:.2f}'
f' (adjusted)')