tiedeman commited on
Commit
e344958
1 Parent(s): 3828d8f

Initial commit

Browse files
.gitattributes CHANGED
@@ -26,3 +26,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
26
  *.zip filter=lfs diff=lfs merge=lfs -text
27
  *.zstandard filter=lfs diff=lfs merge=lfs -text
28
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
26
  *.zip filter=lfs diff=lfs merge=lfs -text
27
  *.zstandard filter=lfs diff=lfs merge=lfs -text
28
  *tfevents* filter=lfs diff=lfs merge=lfs -text
29
+ *.spm filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,336 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - ces+slk
4
+ - cs
5
+ - en
6
+
7
+ tags:
8
+ - translation
9
+
10
+ license: cc-by-4.0
11
+ model-index:
12
+ - name: opus-mt-tc-big-en-ces_slk
13
+ results:
14
+ - task:
15
+ name: Translation eng-ces
16
+ type: translation
17
+ args: eng-ces
18
+ dataset:
19
+ name: flores101-devtest
20
+ type: flores_101
21
+ args: eng ces devtest
22
+ metrics:
23
+ - name: BLEU
24
+ type: bleu
25
+ value: 34.1
26
+ - task:
27
+ name: Translation eng-slk
28
+ type: translation
29
+ args: eng-slk
30
+ dataset:
31
+ name: flores101-devtest
32
+ type: flores_101
33
+ args: eng slk devtest
34
+ metrics:
35
+ - name: BLEU
36
+ type: bleu
37
+ value: 35.9
38
+ - task:
39
+ name: Translation eng-ces
40
+ type: translation
41
+ args: eng-ces
42
+ dataset:
43
+ name: multi30k_test_2016_flickr
44
+ type: multi30k-2016_flickr
45
+ args: eng-ces
46
+ metrics:
47
+ - name: BLEU
48
+ type: bleu
49
+ value: 33.4
50
+ - task:
51
+ name: Translation eng-ces
52
+ type: translation
53
+ args: eng-ces
54
+ dataset:
55
+ name: multi30k_test_2018_flickr
56
+ type: multi30k-2018_flickr
57
+ args: eng-ces
58
+ metrics:
59
+ - name: BLEU
60
+ type: bleu
61
+ value: 33.4
62
+ - task:
63
+ name: Translation eng-ces
64
+ type: translation
65
+ args: eng-ces
66
+ dataset:
67
+ name: news-test2008
68
+ type: news-test2008
69
+ args: eng-ces
70
+ metrics:
71
+ - name: BLEU
72
+ type: bleu
73
+ value: 22.8
74
+ - task:
75
+ name: Translation eng-ces
76
+ type: translation
77
+ args: eng-ces
78
+ dataset:
79
+ name: tatoeba-test-v2021-08-07
80
+ type: tatoeba_mt
81
+ args: eng-ces
82
+ metrics:
83
+ - name: BLEU
84
+ type: bleu
85
+ value: 47.5
86
+ - task:
87
+ name: Translation eng-ces
88
+ type: translation
89
+ args: eng-ces
90
+ dataset:
91
+ name: newstest2009
92
+ type: wmt-2009-news
93
+ args: eng-ces
94
+ metrics:
95
+ - name: BLEU
96
+ type: bleu
97
+ value: 24.3
98
+ - task:
99
+ name: Translation eng-ces
100
+ type: translation
101
+ args: eng-ces
102
+ dataset:
103
+ name: newstest2010
104
+ type: wmt-2010-news
105
+ args: eng-ces
106
+ metrics:
107
+ - name: BLEU
108
+ type: bleu
109
+ value: 24.4
110
+ - task:
111
+ name: Translation eng-ces
112
+ type: translation
113
+ args: eng-ces
114
+ dataset:
115
+ name: newstest2011
116
+ type: wmt-2011-news
117
+ args: eng-ces
118
+ metrics:
119
+ - name: BLEU
120
+ type: bleu
121
+ value: 25.5
122
+ - task:
123
+ name: Translation eng-ces
124
+ type: translation
125
+ args: eng-ces
126
+ dataset:
127
+ name: newstest2012
128
+ type: wmt-2012-news
129
+ args: eng-ces
130
+ metrics:
131
+ - name: BLEU
132
+ type: bleu
133
+ value: 22.6
134
+ - task:
135
+ name: Translation eng-ces
136
+ type: translation
137
+ args: eng-ces
138
+ dataset:
139
+ name: newstest2013
140
+ type: wmt-2013-news
141
+ args: eng-ces
142
+ metrics:
143
+ - name: BLEU
144
+ type: bleu
145
+ value: 27.4
146
+ - task:
147
+ name: Translation eng-ces
148
+ type: translation
149
+ args: eng-ces
150
+ dataset:
151
+ name: newstest2014
152
+ type: wmt-2014-news
153
+ args: eng-ces
154
+ metrics:
155
+ - name: BLEU
156
+ type: bleu
157
+ value: 31.4
158
+ - task:
159
+ name: Translation eng-ces
160
+ type: translation
161
+ args: eng-ces
162
+ dataset:
163
+ name: newstest2015
164
+ type: wmt-2015-news
165
+ args: eng-ces
166
+ metrics:
167
+ - name: BLEU
168
+ type: bleu
169
+ value: 27.0
170
+ - task:
171
+ name: Translation eng-ces
172
+ type: translation
173
+ args: eng-ces
174
+ dataset:
175
+ name: newstest2016
176
+ type: wmt-2016-news
177
+ args: eng-ces
178
+ metrics:
179
+ - name: BLEU
180
+ type: bleu
181
+ value: 29.9
182
+ - task:
183
+ name: Translation eng-ces
184
+ type: translation
185
+ args: eng-ces
186
+ dataset:
187
+ name: newstest2017
188
+ type: wmt-2017-news
189
+ args: eng-ces
190
+ metrics:
191
+ - name: BLEU
192
+ type: bleu
193
+ value: 24.9
194
+ - task:
195
+ name: Translation eng-ces
196
+ type: translation
197
+ args: eng-ces
198
+ dataset:
199
+ name: newstest2018
200
+ type: wmt-2018-news
201
+ args: eng-ces
202
+ metrics:
203
+ - name: BLEU
204
+ type: bleu
205
+ value: 24.6
206
+ - task:
207
+ name: Translation eng-ces
208
+ type: translation
209
+ args: eng-ces
210
+ dataset:
211
+ name: newstest2019
212
+ type: wmt-2019-news
213
+ args: eng-ces
214
+ metrics:
215
+ - name: BLEU
216
+ type: bleu
217
+ value: 26.4
218
+ ---
219
+ # opus-mt-tc-big-en-ces_slk
220
+
221
+ Neural machine translation model for translating from English (en) to unknown (ces+slk).
222
+
223
+ This model is part of the [OPUS-MT project](https://github.com/Helsinki-NLP/Opus-MT), an effort to make neural machine translation models widely available and accessible for many languages in the world. All models are originally trained using the amazing framework of [Marian NMT](https://marian-nmt.github.io/), an efficient NMT implementation written in pure C++. The models have been converted to pyTorch using the transformers library by huggingface. Training data is taken from [OPUS](https://opus.nlpl.eu/) and training pipelines use the procedures of [OPUS-MT-train](https://github.com/Helsinki-NLP/Opus-MT-train).
224
+
225
+ * Publications: [OPUS-MT – Building open translation services for the World](https://aclanthology.org/2020.eamt-1.61/) and [The Tatoeba Translation Challenge – Realistic Data Sets for Low Resource and Multilingual MT](https://aclanthology.org/2020.wmt-1.139/) (Please, cite if you use this model.)
226
+
227
+ ```
228
+ @inproceedings{tiedemann-thottingal-2020-opus,
229
+ title = "{OPUS}-{MT} {--} Building open translation services for the World",
230
+ author = {Tiedemann, J{\"o}rg and Thottingal, Santhosh},
231
+ booktitle = "Proceedings of the 22nd Annual Conference of the European Association for Machine Translation",
232
+ month = nov,
233
+ year = "2020",
234
+ address = "Lisboa, Portugal",
235
+ publisher = "European Association for Machine Translation",
236
+ url = "https://aclanthology.org/2020.eamt-1.61",
237
+ pages = "479--480",
238
+ }
239
+
240
+ @inproceedings{tiedemann-2020-tatoeba,
241
+ title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
242
+ author = {Tiedemann, J{\"o}rg},
243
+ booktitle = "Proceedings of the Fifth Conference on Machine Translation",
244
+ month = nov,
245
+ year = "2020",
246
+ address = "Online",
247
+ publisher = "Association for Computational Linguistics",
248
+ url = "https://aclanthology.org/2020.wmt-1.139",
249
+ pages = "1174--1182",
250
+ }
251
+ ```
252
+
253
+ ## Model info
254
+
255
+ * Release: 2022-03-13
256
+ * source language(s): eng
257
+ * target language(s): ces
258
+ * model: transformer-big
259
+ * data: opusTCv20210807+bt ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
260
+ * tokenization: SentencePiece (spm32k,spm32k)
261
+ * original model: [opusTCv20210807+bt_transformer-big_2022-03-13.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-ces+slk/opusTCv20210807+bt_transformer-big_2022-03-13.zip)
262
+ * more information released models: [OPUS-MT eng-ces+slk README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/eng-ces+slk/README.md)
263
+
264
+ ## Usage
265
+
266
+ A short example code:
267
+
268
+ ```python
269
+ from transformers import MarianMTModel, MarianTokenizer
270
+
271
+ src_text = [
272
+ ">>ces<< We were enemies.",
273
+ ">>ces<< Do you think Tom knows what's going on?"
274
+ ]
275
+
276
+ model_name = "pytorch-models/opus-mt-tc-big-en-ces_slk"
277
+ tokenizer = MarianTokenizer.from_pretrained(model_name)
278
+ model = MarianMTModel.from_pretrained(model_name)
279
+ translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))
280
+
281
+ for t in translated:
282
+ print( tokenizer.decode(t, skip_special_tokens=True) )
283
+
284
+ # expected output:
285
+ # Byli jsme nepřátelé.
286
+ # Myslíš, že Tom ví, co se děje?
287
+ ```
288
+
289
+ You can also use OPUS-MT models with the transformers pipelines, for example:
290
+
291
+ ```python
292
+ from transformers import pipeline
293
+ pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-en-ces_slk")
294
+ print(pipe(">>ces<< We were enemies."))
295
+
296
+ # expected output: Byli jsme nepřátelé.
297
+ ```
298
+
299
+ ## Benchmarks
300
+
301
+ * test set translations: [opusTCv20210807+bt_transformer-big_2022-03-13.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-ces+slk/opusTCv20210807+bt_transformer-big_2022-03-13.test.txt)
302
+ * test set scores: [opusTCv20210807+bt_transformer-big_2022-03-13.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-ces+slk/opusTCv20210807+bt_transformer-big_2022-03-13.eval.txt)
303
+ * benchmark results: [benchmark_results.txt](benchmark_results.txt)
304
+ * benchmark output: [benchmark_translations.zip](benchmark_translations.zip)
305
+
306
+ | langpair | testset | chr-F | BLEU | #sent | #words |
307
+ |----------|---------|-------|-------|-------|--------|
308
+ | eng-ces | tatoeba-test-v2021-08-07 | 0.66128 | 47.5 | 13824 | 91332 |
309
+ | eng-ces | flores101-devtest | 0.60411 | 34.1 | 1012 | 22101 |
310
+ | eng-slk | flores101-devtest | 0.62415 | 35.9 | 1012 | 22543 |
311
+ | eng-ces | multi30k_test_2016_flickr | 0.58547 | 33.4 | 1000 | 10503 |
312
+ | eng-ces | multi30k_test_2018_flickr | 0.59236 | 33.4 | 1071 | 11631 |
313
+ | eng-ces | newssyscomb2009 | 0.52702 | 25.3 | 502 | 10032 |
314
+ | eng-ces | news-test2008 | 0.50286 | 22.8 | 2051 | 42484 |
315
+ | eng-ces | newstest2009 | 0.52152 | 24.3 | 2525 | 55533 |
316
+ | eng-ces | newstest2010 | 0.52527 | 24.4 | 2489 | 52955 |
317
+ | eng-ces | newstest2011 | 0.52721 | 25.5 | 3003 | 65653 |
318
+ | eng-ces | newstest2012 | 0.50007 | 22.6 | 3003 | 65456 |
319
+ | eng-ces | newstest2013 | 0.53643 | 27.4 | 3000 | 57250 |
320
+ | eng-ces | newstest2014 | 0.58944 | 31.4 | 3003 | 59902 |
321
+ | eng-ces | newstest2015 | 0.55094 | 27.0 | 2656 | 45858 |
322
+ | eng-ces | newstest2016 | 0.56864 | 29.9 | 2999 | 56998 |
323
+ | eng-ces | newstest2017 | 0.52504 | 24.9 | 3005 | 54361 |
324
+ | eng-ces | newstest2018 | 0.52490 | 24.6 | 2983 | 54652 |
325
+ | eng-ces | newstest2019 | 0.53994 | 26.4 | 1997 | 43113 |
326
+
327
+ ## Acknowledgements
328
+
329
+ The work is supported by the [European Language Grid](https://www.european-language-grid.eu/) as [pilot project 2866](https://live.european-language-grid.eu/catalogue/#/resource/projects/2866), by the [FoTran project](https://www.helsinki.fi/en/researchgroups/natural-language-understanding-with-cross-lingual-grounding), funded by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 771113), and the [MeMAD project](https://memad.eu/), funded by the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement No 780069. We are also grateful for the generous computational resources and IT infrastructure provided by [CSC -- IT Center for Science](https://www.csc.fi/), Finland.
330
+
331
+ ## Model conversion info
332
+
333
+ * transformers version: 4.16.2
334
+ * OPUS-MT git hash: 3405783
335
+ * port time: Wed Apr 13 16:46:48 EEST 2022
336
+ * port machine: LM0-400-22516.local
benchmark_results.txt ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ eng-ces flores101-dev 0.59502 32.7 997 21183
2
+ eng-slk flores101-dev 0.62025 35.8 997 21796
3
+ eng-ces flores101-devtest 0.60411 34.1 1012 22101
4
+ eng-slk flores101-devtest 0.62415 35.9 1012 22543
5
+ eng-ces multi30k_test_2016_flickr 0.58547 33.4 1000 10503
6
+ eng-ces multi30k_test_2018_flickr 0.59236 33.4 1071 11631
7
+ eng-ces newssyscomb2009 0.52702 25.3 502 10032
8
+ eng-ces news-test2008 0.50286 22.8 2051 42484
9
+ eng-ces newstest2009 0.52152 24.3 2525 55533
10
+ eng-ces newstest2010 0.52527 24.4 2489 52955
11
+ eng-ces newstest2011 0.52721 25.5 3003 65653
12
+ eng-ces newstest2012 0.50007 22.6 3003 65456
13
+ eng-ces newstest2013 0.53643 27.4 3000 57250
14
+ eng-ces newstest2014 0.58944 31.4 3003 59902
15
+ eng-ces newstest2015 0.55094 27.0 2656 45858
16
+ eng-ces newstest2016 0.56864 29.9 2999 56998
17
+ eng-ces newstest2017 0.52504 24.9 3005 54361
18
+ eng-ces newstest2018 0.52490 24.6 2983 54652
19
+ eng-ces newstest2019 0.53994 26.4 1997 43113
20
+ eng-ces tatoeba-test-v2020-07-28 0.66202 47.5 10000 65288
21
+ eng-ces tatoeba-test-v2021-03-30 0.66216 47.6 12076 79375
22
+ eng-ces tatoeba-test-v2021-08-07 0.66128 47.5 13824 91332
benchmark_translations.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75317a7169d824615e52cd68d4ac4046e8fbb2154e9a3eb3bcb784ec859090e1
3
+ size 6591097
config.json ADDED
@@ -0,0 +1,45 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_dropout": 0.0,
3
+ "activation_function": "relu",
4
+ "architectures": [
5
+ "MarianMTModel"
6
+ ],
7
+ "attention_dropout": 0.0,
8
+ "bad_words_ids": [
9
+ [
10
+ 57702
11
+ ]
12
+ ],
13
+ "bos_token_id": 0,
14
+ "classifier_dropout": 0.0,
15
+ "d_model": 1024,
16
+ "decoder_attention_heads": 16,
17
+ "decoder_ffn_dim": 4096,
18
+ "decoder_layerdrop": 0.0,
19
+ "decoder_layers": 6,
20
+ "decoder_start_token_id": 57702,
21
+ "decoder_vocab_size": 57703,
22
+ "dropout": 0.1,
23
+ "encoder_attention_heads": 16,
24
+ "encoder_ffn_dim": 4096,
25
+ "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 6,
27
+ "eos_token_id": 42432,
28
+ "forced_eos_token_id": 42432,
29
+ "init_std": 0.02,
30
+ "is_encoder_decoder": true,
31
+ "max_length": 512,
32
+ "max_position_embeddings": 1024,
33
+ "model_type": "marian",
34
+ "normalize_embedding": false,
35
+ "num_beams": 4,
36
+ "num_hidden_layers": 6,
37
+ "pad_token_id": 57702,
38
+ "scale_embedding": true,
39
+ "share_encoder_decoder_embeddings": true,
40
+ "static_position_embeddings": true,
41
+ "torch_dtype": "float16",
42
+ "transformers_version": "4.18.0.dev0",
43
+ "use_cache": true,
44
+ "vocab_size": 57703
45
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a0c7c6127e1fb1b642191a2c33fd62c8f9b5c9c730a71e1bca073b3c3f37266b
3
+ size 589273347
source.spm ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:adabb94ecde18c3b7d9af17fc922f206cae6f880d8cd6df24b6b629594f5dd29
3
+ size 796504
special_tokens_map.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}
target.spm ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03d425403f18a5e6f1ce7513643e01755341532c88ed26831edaba60fc6043f0
3
+ size 823079
tokenizer_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"source_lang": "en", "target_lang": "ces+slk", "unk_token": "<unk>", "eos_token": "</s>", "pad_token": "<pad>", "model_max_length": 512, "sp_model_kwargs": {}, "separate_vocabs": false, "special_tokens_map_file": null, "name_or_path": "marian-models/opusTCv20210807+bt_transformer-big_2022-03-13/en-ces+slk", "tokenizer_class": "MarianTokenizer"}
vocab.json ADDED
The diff for this file is too large to render. See raw diff