/
models.html
571 lines (533 loc) · 29.9 KB
/
models.html
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
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
<meta name="generator" content="pdoc 0.10.0" />
<title>ktrain.text.ner.models API documentation</title>
<meta name="description" content="" />
<link rel="preload stylesheet" as="style" href="https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/11.0.1/sanitize.min.css" integrity="sha256-PK9q560IAAa6WVRRh76LtCaI8pjTJ2z11v0miyNNjrs=" crossorigin>
<link rel="preload stylesheet" as="style" href="https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/11.0.1/typography.min.css" integrity="sha256-7l/o7C8jubJiy74VsKTidCy1yBkRtiUGbVkYBylBqUg=" crossorigin>
<link rel="stylesheet preload" as="style" href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/10.1.1/styles/github.min.css" crossorigin>
<style>:root{--highlight-color:#fe9}.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:30px;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:1em 0 .50em 0}h3{font-size:1.4em;margin:25px 0 10px 0}h4{margin:0;font-size:105%}h1:target,h2:target,h3:target,h4:target,h5:target,h6:target{background:var(--highlight-color);padding:.2em 0}a{color:#058;text-decoration:none;transition:color .3s ease-in-out}a:hover{color:#e82}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900}pre code{background:#f8f8f8;font-size:.8em;line-height:1.4em}code{background:#f2f2f1;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{background:#f8f8f8;border:0;border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0;padding:1ex}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-weight:bold;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}dt:target .name{background:var(--highlight-color)}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}td{padding:0 .5em}.admonition{padding:.1em .5em;margin-bottom:1em}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.item .name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul{padding-left:1.5em}.toc > ul > li{margin-top:.5em}}</style>
<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
<script defer src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/10.1.1/highlight.min.js" integrity="sha256-Uv3H6lx7dJmRfRvH8TH6kJD1TSK1aFcwgx+mdg3epi8=" crossorigin></script>
<script>window.addEventListener('DOMContentLoaded', () => hljs.initHighlighting())</script>
</head>
<body>
<main>
<article id="content">
<header>
<h1 class="title">Module <code>ktrain.text.ner.models</code></h1>
</header>
<section id="section-intro">
<details class="source">
<summary>
<span>Expand source code</span>
</summary>
<pre><code class="python">from ... import utils as U
from ...imports import *
from . import preprocessor as pp
BILSTM_CRF = "bilstm-crf"
BILSTM = "bilstm"
BILSTM_ELMO = "bilstm-elmo"
BILSTM_CRF_ELMO = "bilstm-crf-elmo"
BILSTM_TRANSFORMER = "bilstm-transformer"
SEQUENCE_TAGGERS = {
BILSTM: "Bidirectional LSTM (https://arxiv.org/abs/1603.01360)",
BILSTM_TRANSFORMER: "Bidirectional LSTM w/ transformer embeddings (multlingual BERT is default)",
BILSTM_CRF: "Bidirectional LSTM-CRF (https://arxiv.org/abs/1603.01360)",
BILSTM_ELMO: "Bidirectional LSTM w/ Elmo embeddings [English only]",
BILSTM_CRF_ELMO: "Bidirectional LSTM-CRF w/ Elmo embeddings [English only]",
}
V1_ONLY_MODELS = [BILSTM_CRF, BILSTM_CRF_ELMO]
TRANSFORMER_MODELS = [BILSTM_TRANSFORMER]
ELMO_MODELS = [BILSTM_ELMO, BILSTM_CRF_ELMO]
def print_sequence_taggers():
for k, v in SEQUENCE_TAGGERS.items():
print("%s: %s" % (k, v))
def sequence_tagger(
name,
preproc,
wv_path_or_url=None,
transformer_model="bert-base-multilingual-cased",
transformer_layers_to_use=U.DEFAULT_TRANSFORMER_LAYERS,
bert_model=None,
word_embedding_dim=100,
char_embedding_dim=25,
word_lstm_size=100,
char_lstm_size=25,
fc_dim=100,
dropout=0.5,
verbose=1,
):
"""
Build and return a sequence tagger (i.e., named entity recognizer).
Args:
name (string): one of:
- 'bilstm-crf' for Bidirectional LSTM-CRF model
- 'bilstm' for Bidirectional LSTM (no CRF layer)
preproc(NERPreprocessor): an instance of NERPreprocessor
wv_path_or_url(str): either a URL or file path toa fasttext word vector file (.vec or .vec.zip or .vec.gz)
Example valid values for wv_path_or_url:
Randomly-initialized word embeeddings:
set wv_path_or_url=None
English pretrained word vectors:
https://dl.fbaipublicfiles.com/fasttext/vectors-english/crawl-300d-2M.vec.zip
Chinese pretrained word vectors:
https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.zh.300.vec.gz
Russian pretrained word vectors:
https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.ru.300.vec.gz
Dutch pretrained word vectors:
https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.nl.300.vec.gz
See these two Web pages for a full list of URLs to word vector files for
different languages:
1. https://fasttext.cc/docs/en/english-vectors.html (for English)
2. https://fasttext.cc/docs/en/crawl-vectors.html (for non-English langages)
Default:None (randomly-initialized word embeddings are used)
transformer_model_name(str): the name of the transformer model. default: 'bert-base-multilingual-cased'
This parameter is only used if bilstm-transformer is selected for name parameter.
The value of this parameter is a name of transformer model from here:
https://huggingface.co/transformers/pretrained_models.html
or a community-uploaded BERT model from here:
https://huggingface.co/models
Example values:
bert-base-multilingual-cased: Multilingual BERT (157 languages) - this is the default
bert-base-cased: English BERT
bert-base-chinese: Chinese BERT
distilbert-base-german-cased: German DistilBert
albert-base-v2: English ALBERT model
monologg/biobert_v1.1_pubmed: community uploaded BioBERT (pretrained on PubMed)
transformer_layers_to_use(list): indices of hidden layers to use. default:[-2] # second-to-last layer
To use the concatenation of last 4 layers: use [-1, -2, -3, -4]
bert_model(str): alias for transformer_model
word_embedding_dim (int): word embedding dimensions.
char_embedding_dim (int): character embedding dimensions.
word_lstm_size (int): character LSTM feature extractor output dimensions.
char_lstm_size (int): word tagger LSTM output dimensions.
fc_dim (int): output fully-connected layer size.
dropout (float): dropout rate.
verbose (boolean): verbosity of output
Return:
model (Model): A Keras Model instance
"""
# backwards compatibility
name = BILSTM_TRANSFORMER if name == "bilstm-bert" else name
if bert_model is not None:
transformer_model = bert_model
warnings.warn(
"The bert_model argument is deprecated - please use transformer_model instead.",
DeprecationWarning,
stacklevel=2,
)
if name not in SEQUENCE_TAGGERS:
raise ValueError(
f"Invalid model name {name}. {'Did you mean bilstm-transformer?' if name == 'bilstm-bert' else ''}"
)
# check BERT
if name in TRANSFORMER_MODELS and not transformer_model:
raise ValueError(
f"transformer_model is required for {BILSTM_TRANSFORMER} models"
)
if name in TRANSFORMER_MODELS and DISABLE_V2_BEHAVIOR:
raise ValueError(
"BERT and other transformer models cannot be used with DISABLE_v2_BEHAVIOR"
)
# check CRF
if not DISABLE_V2_BEHAVIOR and name in V1_ONLY_MODELS:
warnings.warn(
"Falling back to BiLSTM (no CRF) because DISABLE_V2_BEHAVIOR=False"
)
msg = (
"\nIMPORTANT NOTE: ktrain uses the CRF module from keras_contrib, which is not yet\n"
+ "fully compatible with TensorFlow 2. You can still use the BiLSTM-CRF model\n"
+ "in ktrain for sequence tagging with TensorFlow 2, but you must add the\n"
+ "following to the top of your script or notebook BEFORE you import ktrain:\n\n"
+ "import os\n"
+ "os.environ['DISABLE_V2_BEHAVIOR'] = '1'\n\n"
+ "For this run, a vanilla BiLSTM model (with no CRF layer) will be used.\n"
)
print(msg)
name = BILSTM if name == BILSTM_CRF else BILSTM_ELMO
# check for use_char=True
if not DISABLE_V2_BEHAVIOR and preproc.p._use_char:
# turn off masking due to open TF2 issue ##33148: https://github.com/tensorflow/tensorflow/issues/33148
warnings.warn(
"Setting use_char=False: character embeddings cannot be used in TF2 due to open TensorFlow 2 bug (#33148).\n"
+ 'Add os.environ["DISABLE_V2_BEHAVIOR"] = "1" to the top of script if you really want to use it.'
)
preproc.p._use_char = False
if verbose:
emb_names = []
if wv_path_or_url is not None:
emb_names.append(
"word embeddings initialized with fasttext word vectors (%s)"
% (os.path.basename(wv_path_or_url))
)
else:
emb_names.append("word embeddings initialized randomly")
if name in TRANSFORMER_MODELS:
emb_names.append("transformer embeddings with " + transformer_model)
if name in ELMO_MODELS:
emb_names.append("Elmo embeddings for English")
if preproc.p._use_char:
emb_names.append("character embeddings")
if len(emb_names) > 1:
print("Embedding schemes employed (combined with concatenation):")
else:
print("embedding schemes employed:")
for emb_name in emb_names:
print("\t%s" % (emb_name))
print()
# setup embedding
if wv_path_or_url is not None:
wv_model, word_embedding_dim = preproc.get_wv_model(
wv_path_or_url, verbose=verbose
)
else:
wv_model = None
if name == BILSTM_CRF:
use_crf = False if not DISABLE_V2_BEHAVIOR else True # fallback to bilstm
elif name == BILSTM_CRF_ELMO:
use_crf = False if not DISABLE_V2_BEHAVIOR else True # fallback to bilstm
preproc.p.activate_elmo()
elif name == BILSTM:
use_crf = False
elif name == BILSTM_ELMO:
use_crf = False
preproc.p.activate_elmo()
elif name == BILSTM_TRANSFORMER:
use_crf = False
preproc.p.activate_transformer(
transformer_model, layers=transformer_layers_to_use, force=True
)
else:
raise ValueError("Unsupported model name")
from .anago.models import BiLSTMCRF
model = BiLSTMCRF(
char_embedding_dim=char_embedding_dim,
word_embedding_dim=word_embedding_dim,
char_lstm_size=char_lstm_size,
word_lstm_size=word_lstm_size,
fc_dim=fc_dim,
char_vocab_size=preproc.p.char_vocab_size,
word_vocab_size=preproc.p.word_vocab_size,
num_labels=preproc.p.label_size,
dropout=dropout,
use_crf=use_crf,
use_char=preproc.p._use_char,
embeddings=wv_model,
use_elmo=preproc.p.elmo_is_activated(),
use_transformer_with_dim=preproc.p.get_transformer_dim(),
)
model, loss = model.build()
model.compile(loss=loss, optimizer=U.DEFAULT_OPT)
return model</code></pre>
</details>
</section>
<section>
</section>
<section>
</section>
<section>
<h2 class="section-title" id="header-functions">Functions</h2>
<dl>
<dt id="ktrain.text.ner.models.print_sequence_taggers"><code class="name flex">
<span>def <span class="ident">print_sequence_taggers</span></span>(<span>)</span>
</code></dt>
<dd>
<div class="desc"></div>
<details class="source">
<summary>
<span>Expand source code</span>
</summary>
<pre><code class="python">def print_sequence_taggers():
for k, v in SEQUENCE_TAGGERS.items():
print("%s: %s" % (k, v))</code></pre>
</details>
</dd>
<dt id="ktrain.text.ner.models.sequence_tagger"><code class="name flex">
<span>def <span class="ident">sequence_tagger</span></span>(<span>name, preproc, wv_path_or_url=None, transformer_model='bert-base-multilingual-cased', transformer_layers_to_use=[-2], bert_model=None, word_embedding_dim=100, char_embedding_dim=25, word_lstm_size=100, char_lstm_size=25, fc_dim=100, dropout=0.5, verbose=1)</span>
</code></dt>
<dd>
<div class="desc"><p>Build and return a sequence tagger (i.e., named entity recognizer).</p>
<h2 id="args">Args</h2>
<dl>
<dt><strong><code>name</code></strong> : <code>string</code></dt>
<dd>one of:
- 'bilstm-crf' for Bidirectional LSTM-CRF model
- 'bilstm' for Bidirectional LSTM (no CRF layer)</dd>
</dl>
<p>preproc(NERPreprocessor):
an instance of NERPreprocessor
wv_path_or_url(str): either a URL or file path toa fasttext word vector file (.vec or .vec.zip or .vec.gz)
Example valid values for wv_path_or_url:</p>
<pre><code> Randomly-initialized word embeeddings:
set wv_path_or_url=None
English pretrained word vectors:
<https://dl.fbaipublicfiles.com/fasttext/vectors-english/crawl-300d-2M.vec.zip>
Chinese pretrained word vectors:
<https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.zh.300.vec.gz>
Russian pretrained word vectors:
<https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.ru.300.vec.gz>
Dutch pretrained word vectors:
<https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.nl.300.vec.gz>
See these two Web pages for a full list of URLs to word vector files for
different languages:
1. <https://fasttext.cc/docs/en/english-vectors.html> (for English)
2. <https://fasttext.cc/docs/en/crawl-vectors.html> (for non-English langages)
Default:None (randomly-initialized word embeddings are used)
</code></pre>
<p>transformer_model_name(str):
the name of the transformer model.
default: 'bert-base-multilingual-cased'
This parameter is only used if bilstm-transformer is selected for name parameter.
The value of this parameter is a name of transformer model from here:
<a href="https://huggingface.co/transformers/pretrained_models.html">https://huggingface.co/transformers/pretrained_models.html</a>
or a community-uploaded BERT model from here:
<a href="https://huggingface.co/models">https://huggingface.co/models</a>
Example values:
bert-base-multilingual-cased:
Multilingual BERT (157 languages) - this is the default
bert-base-cased:
English BERT
bert-base-chinese: Chinese BERT
distilbert-base-german-cased: German DistilBert
albert-base-v2: English ALBERT model
monologg/biobert_v1.1_pubmed: community uploaded BioBERT (pretrained on PubMed)</p>
<dl>
<dt>transformer_layers_to_use(list): indices of hidden layers to use.
default:[-2] # second-to-last layer</dt>
<dt>To use the concatenation of last 4 layers: use [-1, -2, -3, -4]</dt>
<dt>bert_model(str): alias for transformer_model</dt>
<dt><strong><code>word_embedding_dim</code></strong> : <code>int</code></dt>
<dd>word embedding dimensions.</dd>
<dt><strong><code>char_embedding_dim</code></strong> : <code>int</code></dt>
<dd>character embedding dimensions.</dd>
<dt><strong><code>word_lstm_size</code></strong> : <code>int</code></dt>
<dd>character LSTM feature extractor output dimensions.</dd>
<dt><strong><code>char_lstm_size</code></strong> : <code>int</code></dt>
<dd>word tagger LSTM output dimensions.</dd>
<dt><strong><code>fc_dim</code></strong> : <code>int</code></dt>
<dd>output fully-connected layer size.</dd>
<dt><strong><code>dropout</code></strong> : <code>float</code></dt>
<dd>dropout rate.</dd>
<dt><strong><code>verbose</code></strong> : <code>boolean</code></dt>
<dd>verbosity of output</dd>
</dl>
<h2 id="return">Return</h2>
<p>model (Model): A Keras Model instance</p></div>
<details class="source">
<summary>
<span>Expand source code</span>
</summary>
<pre><code class="python">def sequence_tagger(
name,
preproc,
wv_path_or_url=None,
transformer_model="bert-base-multilingual-cased",
transformer_layers_to_use=U.DEFAULT_TRANSFORMER_LAYERS,
bert_model=None,
word_embedding_dim=100,
char_embedding_dim=25,
word_lstm_size=100,
char_lstm_size=25,
fc_dim=100,
dropout=0.5,
verbose=1,
):
"""
Build and return a sequence tagger (i.e., named entity recognizer).
Args:
name (string): one of:
- 'bilstm-crf' for Bidirectional LSTM-CRF model
- 'bilstm' for Bidirectional LSTM (no CRF layer)
preproc(NERPreprocessor): an instance of NERPreprocessor
wv_path_or_url(str): either a URL or file path toa fasttext word vector file (.vec or .vec.zip or .vec.gz)
Example valid values for wv_path_or_url:
Randomly-initialized word embeeddings:
set wv_path_or_url=None
English pretrained word vectors:
https://dl.fbaipublicfiles.com/fasttext/vectors-english/crawl-300d-2M.vec.zip
Chinese pretrained word vectors:
https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.zh.300.vec.gz
Russian pretrained word vectors:
https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.ru.300.vec.gz
Dutch pretrained word vectors:
https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.nl.300.vec.gz
See these two Web pages for a full list of URLs to word vector files for
different languages:
1. https://fasttext.cc/docs/en/english-vectors.html (for English)
2. https://fasttext.cc/docs/en/crawl-vectors.html (for non-English langages)
Default:None (randomly-initialized word embeddings are used)
transformer_model_name(str): the name of the transformer model. default: 'bert-base-multilingual-cased'
This parameter is only used if bilstm-transformer is selected for name parameter.
The value of this parameter is a name of transformer model from here:
https://huggingface.co/transformers/pretrained_models.html
or a community-uploaded BERT model from here:
https://huggingface.co/models
Example values:
bert-base-multilingual-cased: Multilingual BERT (157 languages) - this is the default
bert-base-cased: English BERT
bert-base-chinese: Chinese BERT
distilbert-base-german-cased: German DistilBert
albert-base-v2: English ALBERT model
monologg/biobert_v1.1_pubmed: community uploaded BioBERT (pretrained on PubMed)
transformer_layers_to_use(list): indices of hidden layers to use. default:[-2] # second-to-last layer
To use the concatenation of last 4 layers: use [-1, -2, -3, -4]
bert_model(str): alias for transformer_model
word_embedding_dim (int): word embedding dimensions.
char_embedding_dim (int): character embedding dimensions.
word_lstm_size (int): character LSTM feature extractor output dimensions.
char_lstm_size (int): word tagger LSTM output dimensions.
fc_dim (int): output fully-connected layer size.
dropout (float): dropout rate.
verbose (boolean): verbosity of output
Return:
model (Model): A Keras Model instance
"""
# backwards compatibility
name = BILSTM_TRANSFORMER if name == "bilstm-bert" else name
if bert_model is not None:
transformer_model = bert_model
warnings.warn(
"The bert_model argument is deprecated - please use transformer_model instead.",
DeprecationWarning,
stacklevel=2,
)
if name not in SEQUENCE_TAGGERS:
raise ValueError(
f"Invalid model name {name}. {'Did you mean bilstm-transformer?' if name == 'bilstm-bert' else ''}"
)
# check BERT
if name in TRANSFORMER_MODELS and not transformer_model:
raise ValueError(
f"transformer_model is required for {BILSTM_TRANSFORMER} models"
)
if name in TRANSFORMER_MODELS and DISABLE_V2_BEHAVIOR:
raise ValueError(
"BERT and other transformer models cannot be used with DISABLE_v2_BEHAVIOR"
)
# check CRF
if not DISABLE_V2_BEHAVIOR and name in V1_ONLY_MODELS:
warnings.warn(
"Falling back to BiLSTM (no CRF) because DISABLE_V2_BEHAVIOR=False"
)
msg = (
"\nIMPORTANT NOTE: ktrain uses the CRF module from keras_contrib, which is not yet\n"
+ "fully compatible with TensorFlow 2. You can still use the BiLSTM-CRF model\n"
+ "in ktrain for sequence tagging with TensorFlow 2, but you must add the\n"
+ "following to the top of your script or notebook BEFORE you import ktrain:\n\n"
+ "import os\n"
+ "os.environ['DISABLE_V2_BEHAVIOR'] = '1'\n\n"
+ "For this run, a vanilla BiLSTM model (with no CRF layer) will be used.\n"
)
print(msg)
name = BILSTM if name == BILSTM_CRF else BILSTM_ELMO
# check for use_char=True
if not DISABLE_V2_BEHAVIOR and preproc.p._use_char:
# turn off masking due to open TF2 issue ##33148: https://github.com/tensorflow/tensorflow/issues/33148
warnings.warn(
"Setting use_char=False: character embeddings cannot be used in TF2 due to open TensorFlow 2 bug (#33148).\n"
+ 'Add os.environ["DISABLE_V2_BEHAVIOR"] = "1" to the top of script if you really want to use it.'
)
preproc.p._use_char = False
if verbose:
emb_names = []
if wv_path_or_url is not None:
emb_names.append(
"word embeddings initialized with fasttext word vectors (%s)"
% (os.path.basename(wv_path_or_url))
)
else:
emb_names.append("word embeddings initialized randomly")
if name in TRANSFORMER_MODELS:
emb_names.append("transformer embeddings with " + transformer_model)
if name in ELMO_MODELS:
emb_names.append("Elmo embeddings for English")
if preproc.p._use_char:
emb_names.append("character embeddings")
if len(emb_names) > 1:
print("Embedding schemes employed (combined with concatenation):")
else:
print("embedding schemes employed:")
for emb_name in emb_names:
print("\t%s" % (emb_name))
print()
# setup embedding
if wv_path_or_url is not None:
wv_model, word_embedding_dim = preproc.get_wv_model(
wv_path_or_url, verbose=verbose
)
else:
wv_model = None
if name == BILSTM_CRF:
use_crf = False if not DISABLE_V2_BEHAVIOR else True # fallback to bilstm
elif name == BILSTM_CRF_ELMO:
use_crf = False if not DISABLE_V2_BEHAVIOR else True # fallback to bilstm
preproc.p.activate_elmo()
elif name == BILSTM:
use_crf = False
elif name == BILSTM_ELMO:
use_crf = False
preproc.p.activate_elmo()
elif name == BILSTM_TRANSFORMER:
use_crf = False
preproc.p.activate_transformer(
transformer_model, layers=transformer_layers_to_use, force=True
)
else:
raise ValueError("Unsupported model name")
from .anago.models import BiLSTMCRF
model = BiLSTMCRF(
char_embedding_dim=char_embedding_dim,
word_embedding_dim=word_embedding_dim,
char_lstm_size=char_lstm_size,
word_lstm_size=word_lstm_size,
fc_dim=fc_dim,
char_vocab_size=preproc.p.char_vocab_size,
word_vocab_size=preproc.p.word_vocab_size,
num_labels=preproc.p.label_size,
dropout=dropout,
use_crf=use_crf,
use_char=preproc.p._use_char,
embeddings=wv_model,
use_elmo=preproc.p.elmo_is_activated(),
use_transformer_with_dim=preproc.p.get_transformer_dim(),
)
model, loss = model.build()
model.compile(loss=loss, optimizer=U.DEFAULT_OPT)
return model</code></pre>
</details>
</dd>
</dl>
</section>
<section>
</section>
</article>
<nav id="sidebar">
<h1>Index</h1>
<div class="toc">
<ul></ul>
</div>
<ul id="index">
<li><h3>Super-module</h3>
<ul>
<li><code><a title="ktrain.text.ner" href="index.html">ktrain.text.ner</a></code></li>
</ul>
</li>
<li><h3><a href="#header-functions">Functions</a></h3>
<ul class="">
<li><code><a title="ktrain.text.ner.models.print_sequence_taggers" href="#ktrain.text.ner.models.print_sequence_taggers">print_sequence_taggers</a></code></li>
<li><code><a title="ktrain.text.ner.models.sequence_tagger" href="#ktrain.text.ner.models.sequence_tagger">sequence_tagger</a></code></li>
</ul>
</li>
</ul>
</nav>
</main>
<footer id="footer">
<p>Generated by <a href="https://pdoc3.github.io/pdoc" title="pdoc: Python API documentation generator"><cite>pdoc</cite> 0.10.0</a>.</p>
</footer>
</body>
</html>