-
Notifications
You must be signed in to change notification settings - Fork 2
/
evaluate.py
170 lines (138 loc) · 6.94 KB
/
evaluate.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
import logging
import os
from collections import namedtuple
from pathlib import Path
import fire
import json
from typing import Union
from sentevalpl.tasks import get_task_names
from utils.analyzer import PolishAnalyzer
from methods.base import EmbeddingBase
from methods.utils import cached
from sentevalpl.engine import SE
root_dir = os.path.dirname(os.path.realpath(__file__))
TaskDef = namedtuple("TaskDef", ("id", "name", ))
class SentEvaluator(object):
def random(self, **kwargs):
from methods.word_vectors import RandomEmbedding
method = RandomEmbedding()
self.evaluate(method, "random", **kwargs)
def word2vec(self, **kwargs):
path: Path = Path(root_dir, "resources/word2vec/word2vec_100_3_polish.bin")
self.evaluate_keyed_vectors(path, "word2vec", **kwargs)
def glove(self, **kwargs):
path: Path = Path(root_dir, "resources/glove/glove_100_3_polish.txt")
self.evaluate_keyed_vectors(path, "glove", **kwargs)
def fasttext(self, **kwargs):
path: Path = Path(root_dir, "resources/fasttext/fasttext_100_3_polish.bin")
self.evaluate_keyed_vectors(path, "fasttext", **kwargs)
def elmo_all(self, **kwargs):
from methods.word_vectors import ElmoEmbedding
method = ElmoEmbedding(Path(root_dir, "resources/elmo/"), layers="all")
self.evaluate(method, "elmo_all", **kwargs)
def elmo_top(self, **kwargs):
from methods.word_vectors import ElmoEmbedding
method = ElmoEmbedding(Path(root_dir, "resources/elmo/"), layers="top")
self.evaluate(method, "elmo_top", **kwargs)
def flair(self, **kwargs):
from methods.word_vectors import FlairEmbedding
method = FlairEmbedding()
self.evaluate(method, "flair", **kwargs)
def bert(self, **kwargs):
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding("bert-base-multilingual-cased", layers="top")
self.evaluate(method, "bert", **kwargs)
def roberta_base_top(self, **kwargs):
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding("resources/roberta_base_transformers/", layers="top")
self.evaluate(method, "roberta_base_top", **kwargs)
def roberta_base_all(self, **kwargs):
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding("resources/roberta_base_transformers/", layers="all")
self.evaluate(method, "roberta_base_all", **kwargs)
def roberta_large_top(self, **kwargs):
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding("resources/roberta_large_transformers/", layers="top")
self.evaluate(method, "roberta_large_top", **kwargs)
def roberta_large_all(self, **kwargs):
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding("resources/roberta_large_transformers/", layers="all")
self.evaluate(method, "roberta_large_all", **kwargs)
def xlmr_base_top(self, **kwargs):
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding("xlm-roberta-base", layers="top")
self.evaluate(method, "xlmr_base_top", **kwargs)
def xlmr_base_all(self, **kwargs):
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding("xlm-roberta-base", layers="all")
self.evaluate(method, "xlmr_base_all", **kwargs)
def xlmr_large_top(self, **kwargs):
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding("xlm-roberta-large", layers="top")
self.evaluate(method, "xlmr_large_top", **kwargs)
def xlmr_large_all(self, **kwargs):
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding("xlm-roberta-large", layers="all")
self.evaluate(method, "xlmr_large_all", **kwargs)
def huggingface(self, **kwargs):
model_name = kwargs.get("model_name")
from methods.huggingface import HuggingfaceModelEmbedding
method = HuggingfaceModelEmbedding(model_name, layers="top")
self.evaluate(method, "huggingface_" + model_name.replace("/", "_"), **kwargs)
def laser(self, **kwargs):
from methods.laser import LaserEmbedding
method = LaserEmbedding()
self.evaluate(method, "laser", **kwargs)
def use(self, **kwargs):
from methods.use import USEEmbedding
method = USEEmbedding()
self.evaluate(method, "use", **kwargs)
def labse(self, **kwargs):
self.sentence_transformers(model_name="sentence-transformers/LaBSE", **kwargs)
def sentence_transformers(self, **kwargs):
from methods.sentence_transformer import SentenceTransformersEmbedding
model_name = kwargs.get("model_name")
method = SentenceTransformersEmbedding(model_name)
self.evaluate(method, "sentence_transformers_" + model_name.replace("/", "_"), **kwargs)
def evaluate_keyed_vectors(self, path: Union[Path, str], name: str, **kwargs):
if isinstance(path, str): path = Path(path)
if kwargs.get("sif"):
from methods.sif import SIFEmbedding
name = name + "_sif"
method = SIFEmbedding(path)
else:
from methods.word_vectors import KeyedVectorsEmbedding
pooling = "avg"
if kwargs.get("pooling") and kwargs.get("pooling") != "avg":
pooling = kwargs.get("pooling")
name = name + "_" +pooling
method = KeyedVectorsEmbedding(path, pooling=pooling)
self.evaluate(method, name, **kwargs)
def evaluate(self, method: EmbeddingBase, method_name: str, **kwargs):
logging.basicConfig(format='%(asctime)s : %(message)s', level=logging.DEBUG)
logging.root.setLevel(logging.DEBUG)
params = {
"task_path": os.path.join(root_dir, "resources"),
"usepytorch": True,
"kfold": 5,
"lemmatize": True,
"batch_size": 512,
"classifier": {"nhid": 50, "optim": "rmsprop", "batch_size": 128, "tenacity": 3, "epoch_size": 10},
"analyzer": PolishAnalyzer()
}
params.update(kwargs)
cache_dir = Path(root_dir, f".cache/{method_name}")
cache_dir.mkdir(parents=True, exist_ok=True)
se = SE(params, cached(method.batcher, cache_dir), method.prepare)
transfer_tasks = get_task_names()
results = se.eval(transfer_tasks)
for key, val in results.items():
if "yhat" in val.keys():
del val["yhat"]
results = {"method": method_name, "results": results}
logging.info(results)
with open(os.path.join(root_dir, "results.txt"), "a+") as output_file:
output_file.write(json.dumps(results))
output_file.write("\n")
if __name__ == '__main__':
fire.Fire(SentEvaluator)