tiedeman commited on
Commit
4cd677d
1 Parent(s): efdb670

Initial commit

Browse files
Files changed (2) hide show
  1. README.md +4 -8
  2. README.md~ +213 -0
README.md CHANGED
@@ -146,15 +146,11 @@ This model is part of the [OPUS-MT project](https://github.com/Helsinki-NLP/Opus
146
  * Release: 2021-12-08
147
  * source language(s): fin
148
  * target language(s): eng
149
- * valid target language labels: >>eng<<
150
  * model: transformer-big (big)
151
  * data: opusTCv20210807+bt ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
152
  * tokenization: SentencePiece (spm32k,spm32k)
153
  * original model: [opusTCv20210807+bt-2021-12-08.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/fin-eng/opusTCv20210807+bt-2021-12-08.zip)
154
  * more information released models: [OPUS-MT fin-eng README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/fin-eng/README.md)
155
- * more information about the model: [MarianMT](https://huggingface.co/docs/transformers/model_doc/marian)
156
-
157
- This is a multilingual translation model with multiple target languages. A sentence initial language token is required in the form of `>>id<<` (id = valid target language ID), e.g. `>>eng<<`
158
 
159
  ## Usage
160
 
@@ -164,8 +160,8 @@ A short example code:
164
  from transformers import MarianMTModel, MarianTokenizer
165
 
166
  src_text = [
167
- ">>eng<< Replace this with text in an accepted source language.",
168
- ">>eng<< This is the second sentence."
169
  ]
170
 
171
  model_name = "pytorch-models/opus-mt-tc-big-fi-en"
@@ -182,7 +178,7 @@ You can also use OPUS-MT models with the transformers pipelines, for example:
182
  ```python
183
  from transformers import pipeline
184
  pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-fi-en")
185
- print(pipe(">>eng<< Replace this with text in an accepted source language."))
186
  ```
187
 
188
  ## Benchmarks
@@ -213,5 +209,5 @@ The work is supported by the [European Language Grid](https://www.european-langu
213
 
214
  * transformers version: 4.16.2
215
  * OPUS-MT git hash: f084bad
216
- * port time: Tue Mar 22 14:44:34 EET 2022
217
  * port machine: LM0-400-22516.local
 
146
  * Release: 2021-12-08
147
  * source language(s): fin
148
  * target language(s): eng
 
149
  * model: transformer-big (big)
150
  * data: opusTCv20210807+bt ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
151
  * tokenization: SentencePiece (spm32k,spm32k)
152
  * original model: [opusTCv20210807+bt-2021-12-08.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/fin-eng/opusTCv20210807+bt-2021-12-08.zip)
153
  * more information released models: [OPUS-MT fin-eng README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/fin-eng/README.md)
 
 
 
154
 
155
  ## Usage
156
 
 
160
  from transformers import MarianMTModel, MarianTokenizer
161
 
162
  src_text = [
163
+ "Kolme kolmanteen on kaksikymmentäseitsemän.",
164
+ "Heille syntyi poikavauva."
165
  ]
166
 
167
  model_name = "pytorch-models/opus-mt-tc-big-fi-en"
 
178
  ```python
179
  from transformers import pipeline
180
  pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-fi-en")
181
+ print(pipe("Kolme kolmanteen on kaksikymmentäseitsemän."))
182
  ```
183
 
184
  ## Benchmarks
 
209
 
210
  * transformers version: 4.16.2
211
  * OPUS-MT git hash: f084bad
212
+ * port time: Tue Mar 22 14:52:19 EET 2022
213
  * port machine: LM0-400-22516.local
README.md~ ADDED
@@ -0,0 +1,213 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ - fi
5
+
6
+ tags:
7
+ - translation
8
+
9
+ license: cc-by-4.0
10
+ model-index:
11
+ - name: opus-mt-tc-big-fi-en
12
+ results:
13
+ - task:
14
+ name: Translation fin-eng
15
+ type: translation
16
+ args: fin-eng
17
+ dataset:
18
+ name: flores101-devtest
19
+ type: flores_101
20
+ args: fin eng devtest
21
+ metrics:
22
+ - name: BLEU
23
+ type: bleu
24
+ value: 35.4
25
+ - task:
26
+ name: Translation fin-eng
27
+ type: translation
28
+ args: fin-eng
29
+ dataset:
30
+ name: newsdev2015
31
+ type: newsdev2015
32
+ args: fin-eng
33
+ metrics:
34
+ - name: BLEU
35
+ type: bleu
36
+ value: 28.6
37
+ - task:
38
+ name: Translation fin-eng
39
+ type: translation
40
+ args: fin-eng
41
+ dataset:
42
+ name: tatoeba-test-v2021-08-07
43
+ type: tatoeba_mt
44
+ args: fin-eng
45
+ metrics:
46
+ - name: BLEU
47
+ type: bleu
48
+ value: 57.4
49
+ - task:
50
+ name: Translation fin-eng
51
+ type: translation
52
+ args: fin-eng
53
+ dataset:
54
+ name: newstest2015
55
+ type: wmt-2015-news
56
+ args: fin-eng
57
+ metrics:
58
+ - name: BLEU
59
+ type: bleu
60
+ value: 29.9
61
+ - task:
62
+ name: Translation fin-eng
63
+ type: translation
64
+ args: fin-eng
65
+ dataset:
66
+ name: newstest2016
67
+ type: wmt-2016-news
68
+ args: fin-eng
69
+ metrics:
70
+ - name: BLEU
71
+ type: bleu
72
+ value: 34.3
73
+ - task:
74
+ name: Translation fin-eng
75
+ type: translation
76
+ args: fin-eng
77
+ dataset:
78
+ name: newstest2017
79
+ type: wmt-2017-news
80
+ args: fin-eng
81
+ metrics:
82
+ - name: BLEU
83
+ type: bleu
84
+ value: 37.3
85
+ - task:
86
+ name: Translation fin-eng
87
+ type: translation
88
+ args: fin-eng
89
+ dataset:
90
+ name: newstest2018
91
+ type: wmt-2018-news
92
+ args: fin-eng
93
+ metrics:
94
+ - name: BLEU
95
+ type: bleu
96
+ value: 27.1
97
+ - task:
98
+ name: Translation fin-eng
99
+ type: translation
100
+ args: fin-eng
101
+ dataset:
102
+ name: newstest2019
103
+ type: wmt-2019-news
104
+ args: fin-eng
105
+ metrics:
106
+ - name: BLEU
107
+ type: bleu
108
+ value: 32.7
109
+ ---
110
+ # opus-mt-tc-big-fi-en
111
+
112
+ Neural machine translation model for translating from Finnish (fi) to English (en).
113
+
114
+ 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).
115
+
116
+ * 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.)
117
+
118
+ ```
119
+ @inproceedings{tiedemann-thottingal-2020-opus,
120
+ title = "{OPUS}-{MT} {--} Building open translation services for the World",
121
+ author = {Tiedemann, J{\"o}rg and Thottingal, Santhosh},
122
+ booktitle = "Proceedings of the 22nd Annual Conference of the European Association for Machine Translation",
123
+ month = nov,
124
+ year = "2020",
125
+ address = "Lisboa, Portugal",
126
+ publisher = "European Association for Machine Translation",
127
+ url = "https://aclanthology.org/2020.eamt-1.61",
128
+ pages = "479--480",
129
+ }
130
+
131
+ @inproceedings{tiedemann-2020-tatoeba,
132
+ title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
133
+ author = {Tiedemann, J{\"o}rg},
134
+ booktitle = "Proceedings of the Fifth Conference on Machine Translation",
135
+ month = nov,
136
+ year = "2020",
137
+ address = "Online",
138
+ publisher = "Association for Computational Linguistics",
139
+ url = "https://aclanthology.org/2020.wmt-1.139",
140
+ pages = "1174--1182",
141
+ }
142
+ ```
143
+
144
+ ## Model info
145
+
146
+ * Release: 2021-12-08
147
+ * source language(s): fin
148
+ * target language(s): eng
149
+ * model: transformer-big (big)
150
+ * data: opusTCv20210807+bt ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
151
+ * tokenization: SentencePiece (spm32k,spm32k)
152
+ * original model: [opusTCv20210807+bt-2021-12-08.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/fin-eng/opusTCv20210807+bt-2021-12-08.zip)
153
+ * more information released models: [OPUS-MT fin-eng README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/fin-eng/README.md)
154
+
155
+ ## Usage
156
+
157
+ A short example code:
158
+
159
+ ```python
160
+ from transformers import MarianMTModel, MarianTokenizer
161
+
162
+ src_text = [
163
+ "Replace this with text in an accepted source language.",
164
+ "This is the second sentence."
165
+ ]
166
+
167
+ model_name = "pytorch-models/opus-mt-tc-big-fi-en"
168
+ tokenizer = MarianTokenizer.from_pretrained(model_name)
169
+ model = MarianMTModel.from_pretrained(model_name)
170
+ translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))
171
+
172
+ for t in translated:
173
+ print( tokenizer.decode(t, skip_special_tokens=True) )
174
+ ```
175
+
176
+ You can also use OPUS-MT models with the transformers pipelines, for example:
177
+
178
+ ```python
179
+ from transformers import pipeline
180
+ pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-fi-en")
181
+ print(pipe("Replace this with text in an accepted source language."))
182
+ ```
183
+
184
+ ## Benchmarks
185
+
186
+ * test set translations: [opusTCv20210807+bt-2021-12-08.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/fin-eng/opusTCv20210807+bt-2021-12-08.test.txt)
187
+ * test set scores: [opusTCv20210807+bt-2021-12-08.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/fin-eng/opusTCv20210807+bt-2021-12-08.eval.txt)
188
+ * benchmark results: [benchmark_results.txt](benchmark_results.txt)
189
+ * benchmark output: [benchmark_translations.zip](benchmark_translations.zip)
190
+
191
+ | langpair | testset | chr-F | BLEU | #sent | #words |
192
+ |----------|---------|-------|-------|-------|--------|
193
+ | fin-eng | tatoeba-test-v2021-08-07 | 0.72298 | 57.4 | 10690 | 80552 |
194
+ | fin-eng | flores101-devtest | 0.62521 | 35.4 | 1012 | 24721 |
195
+ | fin-eng | newsdev2015 | 0.56232 | 28.6 | 1500 | 32012 |
196
+ | fin-eng | newstest2015 | 0.57469 | 29.9 | 1370 | 27270 |
197
+ | fin-eng | newstest2016 | 0.60715 | 34.3 | 3000 | 62945 |
198
+ | fin-eng | newstest2017 | 0.63050 | 37.3 | 3002 | 61846 |
199
+ | fin-eng | newstest2018 | 0.54199 | 27.1 | 3000 | 62325 |
200
+ | fin-eng | newstest2019 | 0.59620 | 32.7 | 1996 | 36215 |
201
+ | fin-eng | newstestB2016 | 0.55472 | 27.9 | 3000 | 62945 |
202
+ | fin-eng | newstestB2017 | 0.58847 | 31.1 | 3002 | 61846 |
203
+
204
+ ## Acknowledgements
205
+
206
+ 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.
207
+
208
+ ## Model conversion info
209
+
210
+ * transformers version: 4.16.2
211
+ * OPUS-MT git hash: f084bad
212
+ * port time: Tue Mar 22 14:52:19 EET 2022
213
+ * port machine: LM0-400-22516.local