-
Notifications
You must be signed in to change notification settings - Fork 9
/
evaluate_test_ppl.py
191 lines (162 loc) · 9.6 KB
/
evaluate_test_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
#(c) 2021 NCSOFT Corporation & Korea University. All rights reserved.
import logging
import random
from argparse import ArgumentParser
from pprint import pformat
from tqdm import tqdm
import torch
import torch.nn.functional as F
from torch.nn import Sigmoid, Softmax, CrossEntropyLoss
from data_utils import get_testdata_loaders, add_special_tokens_
logger = logging.getLogger(__file__)
SPECIAL_TOKENS = ["<machine>", "<human>", "<persona>", "<knowledge>"]
def top_filtering(logits, top_k=0., top_p=0.9, threshold=-float('Inf'), filter_value=-float('Inf')):
""" Filter a distribution of logits using top-k, top-p (nucleus) and/or threshold filtering
Args:
logits: logits distribution shape (vocabulary size)
top_k: <=0: no filtering, >0: keep only top k tokens with highest probability.
top_p: <=0.0: no filtering, >0.0: keep only a subset S of candidates, where S is the smallest subset
whose total probability mass is greater than or equal to the threshold top_p.
In practice, we select the highest probability tokens whose cumulative probability mass exceeds
the threshold top_p.
threshold: a minimal threshold to keep logits
"""
assert logits.dim() == 1 # Only work for batch size 1 for now - could update but it would obfuscate a bit the code
top_k = min(top_k, logits.size(-1))
if top_k > 0:
# Remove all tokens with a probability less than the last token in the top-k tokens
indices_to_remove = logits < torch.topk(logits, top_k)[0][..., -1, None]
logits[indices_to_remove] = filter_value
if top_p > 0.0:
# Compute cumulative probabilities of sorted tokens
sorted_logits, sorted_indices = torch.sort(logits, descending=True)
cumulative_probabilities = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
# Remove tokens with cumulative probability above the threshold
sorted_indices_to_remove = cumulative_probabilities > top_p
# Shift the indices to the right to keep also the first token above the threshold
sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
sorted_indices_to_remove[..., 0] = 0
# Back to unsorted indices and set them to -infinity
indices_to_remove = sorted_indices[sorted_indices_to_remove]
logits[indices_to_remove] = filter_value
indices_to_remove = logits < threshold
logits[indices_to_remove] = filter_value
return logits
def run():
parser = ArgumentParser()
parser.add_argument("--test_dataset_path", type=str, default="data/test_focus.json", help="Path or url of the dataset. If empty download from S3.")
parser.add_argument("--test_dataset_cache", type=str, default='data/focus_cache.tar.gz', help="Path or url of the dataset cache")
parser.add_argument("--model_name", type=str, default="", help="{GPT2, BART, transformer-decoder, transformer-encdec}")
parser.add_argument("--model_checkpoint", type=str, default="", help="Path, url or short name of the model")
parser.add_argument("--max_history", type=int, default=1, help="Number of previous utterances to keep in history")
parser.add_argument("--test_batch_size", type=int, default=1, help="Batch size for testing")
parser.add_argument("--device", type=str, default="cuda" if torch.cuda.is_available() else "cpu", help="Device (cuda or cpu)")
parser.add_argument("--no_sample", action='store_true', help="Set to use greedy decoding instead of sampling")
parser.add_argument("--max_length", type=int, default=20, help="Maximum length of the output utterances")
parser.add_argument("--min_length", type=int, default=1, help="Minimum length of the output utterances")
parser.add_argument("--inference", action='store_true', help="If true, inference with gold knowledge")
parser.add_argument("--seed", type=int, default=19950604, help="Seed")
parser.add_argument("--temperature", type=int, default=0.7, help="Sampling softmax temperature")
parser.add_argument("--top_k", type=int, default=0, help="Filter top-k tokens before sampling (<=0: no filtering)")
parser.add_argument("--top_p", type=float, default=0.9, help="Nucleus filtering (top-p) before sampling (<=0.0: no filtering)")
parser.add_argument("--local_rank", type=int, default=-1, help="Local rank for distributed training (-1: not distributed)")
args = parser.parse_args()
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__file__)
logger.info(pformat(args))
args.distributed = (args.local_rank != -1)
if args.seed != 0:
random.seed(args.seed)
torch.random.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
logger.info("Get model and tokenizer")
if args.model_name == 'GPT2':
from transformers import GPT2Tokenizer
from classification_modules import GPT2PK_ctxt
tokenizer = GPT2Tokenizer.from_pretrained(args.model_checkpoint)
model = GPT2PK_ctxt.from_pretrained(args.model_checkpoint)
model.to(args.device)
add_special_tokens_(model, tokenizer)
elif args.model_name == 'BART':
from transformers import BartTokenizer
from classification_modules import BARTPK_ctxt
tokenizer = BartTokenizer.from_pretrained(args.model_checkpoint)
model = BARTPK_ctxt.from_pretrained(args.model_checkpoint)
model.to(args.device)
add_special_tokens_(model, tokenizer)
elif args.model_name == 'transformer-decoder':
from transformers import GPT2Tokenizer
from classification_modules import GPT2PK_ctxt
tokenizer = GPT2Tokenizer.from_pretrained(args.model_checkpoint)
model = GPT2PK_ctxt.from_pretrained(args.model_checkpoint)
model.to(args.device)
add_special_tokens_(model, tokenizer)
elif args.model_name == 'transformer-encdec':
from transformers import BartTokenizer
from classification_modules import BARTPK_ctxt
tokenizer = BartTokenizer.from_pretrained(args.model_checkpoint)
model = BARTPK_ctxt.from_pretrained(args.model_checkpoint)
model.to(args.device)
add_special_tokens_(model, tokenizer)
else:
raise NotImplementedError
logger.info("Prepare datasets")
test_loader, test_sampler = get_testdata_loaders(args, tokenizer, generation=False)
with torch.no_grad():
ppl_avg = []
for test_data in tqdm(test_loader):
loss_fct = CrossEntropyLoss(ignore_index=-100)
if model.config.model_type == 'gpt2':
input_ids, input_eos, lm_labels, token_type_ids, mc_token_ids, persona_candidates, persona_can_idx, persona_grounding, knowledge_candidates, \
knowledge_can_idx, knowledge_grounding, tot_knowledge, tot_knowledge_token_ids, tot_knowledge_eos, reply, dialog, dialog_tti = test_data
elif model.config.model_type == 'bart':
input_ids, input_eos, decoder_input_ids, lm_labels, token_type_ids, mc_token_ids, persona_candidates, persona_can_idx, persona_grounding, knowledge_candidates, \
knowledge_can_idx, knowledge_grounding, tot_knowledge, tot_knowledge_eos, reply, dialog = test_data
else:
raise NotImplementedError
if model.config.model_type == 'gpt2':
output = model(
input_ids=input_ids,
input_eos=input_eos,
token_type_ids=token_type_ids,
only_dial_input_ids=dialog,
only_dial_token_type_ids=dialog_tti,
persona_input_ids=persona_candidates,
knowledge_input_ids=knowledge_candidates,
persona_can_idx=persona_can_idx,
knowledge_can_idx=knowledge_can_idx,
tot_knowledge=tot_knowledge,
tot_knowledge_token_ids=tot_knowledge_token_ids,
tot_knowledge_eos=tot_knowledge_eos,
training=False,
mc_token_ids=mc_token_ids
)
lm_labels, lm_logits, knowledge_logits, persona_logits = output[0], output[1], output[2], output[3]
lm_logits_flat_shifted = lm_logits[:, :-1, :].contiguous().view(-1, lm_logits.size(-1))
lm_labels_flat_shifted = lm_labels[:, 1:].contiguous().view(-1)
lm_loss = loss_fct(lm_logits_flat_shifted, lm_labels_flat_shifted)
elif model.config.model_type == 'bart':
output = model(
input_ids=input_ids,
input_eos=input_eos,
only_dial_input_ids=dialog,
decoder_input_ids=decoder_input_ids,
persona_input_ids=persona_candidates,
knowledge_input_ids=knowledge_candidates,
persona_can_idx=persona_can_idx,
knowledge_can_idx=knowledge_can_idx,
tot_knowledge=tot_knowledge,
tot_knowledge_eos=tot_knowledge_eos,
training=False,
mc_token_ids=mc_token_ids
)
lm_logits, knowledge_logits, persona_logits = output[0], output[1], output[2]
lm_logits_flat_shifted = lm_logits[:, :-1, :].contiguous().view(-1, lm_logits.size(-1))
lm_labels_flat_shifted = lm_labels[:, 1:].contiguous().view(-1)
lm_loss = loss_fct(lm_logits_flat_shifted, lm_labels_flat_shifted)
#ppl
ppl = torch.exp(lm_loss).item()
ppl_avg.append(ppl)
print("ppl: ", sum(ppl_avg) / len(ppl_avg))
if __name__ == "__main__":
run()