tiedeman commited on
Commit
20c3d21
1 Parent(s): 8c54fda

Initial commit

Browse files
Files changed (4) hide show
  1. README.md +1664 -71
  2. benchmark_translations.zip +3 -0
  3. benchmarks.tsv +199 -0
  4. pytorch_model.bin +1 -1
README.md CHANGED
@@ -1,27 +1,1540 @@
1
  ---
2
  language:
 
 
 
 
3
  - gmw
4
- - gmw
5
- datasets:
6
- - opus
 
 
 
 
 
7
  tags:
8
  - translation
 
9
  license: cc-by-4.0
10
  model-index:
11
  - name: opus-mt-tc-base-gmw-gmw
12
  results:
13
- - task:
14
- name: Translation AFR - DEU
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
15
  type: translation
16
  args: afr-deu
17
  dataset:
18
- name: Tatoeba - Test
19
- type: tatoeba
20
- args: gmw-gmw
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21
  metrics:
22
  - name: BLEU
23
  type: bleu
24
- value: 48.5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25
  ---
26
  # opus-mt-tc-base-gmw-gmw
27
 
@@ -31,6 +1544,32 @@ This model is part of the [OPUS-MT project](https://github.com/Helsinki-NLP/Opus
31
 
32
  * 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.)
33
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
34
  ## Model info
35
 
36
  * Release: 2021-02-23
@@ -41,82 +1580,136 @@ This model is part of the [OPUS-MT project](https://github.com/Helsinki-NLP/Opus
41
  * data: opus ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
42
  * tokenization: SentencePiece (spm32k,spm32k)
43
  * original model: [opus-2021-02-23.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/gmw-gmw/opus-2021-02-23.zip)
44
- * more information: [OPUS-MT models (gmw-gmw)](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/gmw-gmw/README.md)
 
45
 
46
  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. `>>afr<<`
47
 
48
  ## Usage
49
 
50
- You can use OPUS-MT models with the transformers pipelines, for example:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
51
 
52
  ```python
53
  from transformers import pipeline
54
  pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-base-gmw-gmw")
55
- print(pipe(">>afr<< Replace this with text in an accepted source language."))
 
 
56
  ```
57
 
58
  ## Benchmarks
59
 
60
  * test set translations: [opus-2021-02-23.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/gmw-gmw/opus-2021-02-23.test.txt)
61
  * test set scores: [opus-2021-02-23.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/gmw-gmw/opus-2021-02-23.eval.txt)
 
 
62
 
63
- | langpair | testset | BLEU | chr-F | #sent | #words | BP |
64
- |----------|---------|-------|-------|-------|--------|----|
65
- | afr-deu | Tatoeba-test | 48.5 | 0.677 | 1583 | 9105 | 1.000 |
66
- | afr-eng | Tatoeba-test | 58.7 | 0.727 | 1374 | 9622 | 0.995 |
67
- | afr-nld | Tatoeba-test | 54.7 | 0.713 | 1056 | 6710 | 0.989 |
68
- | deu-afr | Tatoeba-test | 52.4 | 0.697 | 1583 | 9507 | 1.000 |
69
- | deu-eng | newssyscomb2009 | 25.4 | 0.527 | 502 | 11821 | 0.986 |
70
- | deu-eng | news-test2008 | 23.9 | 0.519 | 2051 | 49380 | 0.992 |
71
- | deu-eng | newstest2009 | 23.5 | 0.517 | 2525 | 65402 | 0.978 |
72
- | deu-eng | newstest2010 | 26.1 | 0.548 | 2489 | 61724 | 1.000 |
73
- | deu-eng | newstest2011 | 23.9 | 0.525 | 3003 | 74681 | 1.000 |
74
- | deu-eng | newstest2012 | 25.0 | 0.533 | 3003 | 72812 | 1.000 |
75
- | deu-eng | newstest2013 | 27.7 | 0.549 | 3000 | 64505 | 1.000 |
76
- | deu-eng | newstest2014-deen | 27.4 | 0.549 | 3003 | 67337 | 0.977 |
77
- | deu-eng | newstest2015-ende | 28.8 | 0.554 | 2169 | 46443 | 0.973 |
78
- | deu-eng | newstest2016-ende | 33.7 | 0.598 | 2999 | 64126 | 1.000 |
79
- | deu-eng | newstest2017-ende | 29.6 | 0.562 | 3004 | 64399 | 0.979 |
80
- | deu-eng | newstest2018-ende | 36.3 | 0.611 | 2998 | 67013 | 0.977 |
81
- | deu-eng | newstest2019-deen | 32.7 | 0.585 | 2000 | 39282 | 0.984 |
82
- | deu-eng | Tatoeba-test | 44.7 | 0.629 | 10000 | 81233 | 0.975 |
83
- | deu-nds | Tatoeba-test | 18.7 | 0.444 | 10000 | 76144 | 0.988 |
84
- | deu-nld | Tatoeba-test | 48.7 | 0.672 | 10000 | 73546 | 0.969 |
85
- | eng-afr | Tatoeba-test | 56.5 | 0.735 | 1374 | 10317 | 0.984 |
86
- | eng-deu | newssyscomb2009 | 19.4 | 0.503 | 502 | 11271 | 0.991 |
87
- | eng-deu | news-test2008 | 19.5 | 0.493 | 2051 | 47427 | 0.996 |
88
- | eng-deu | newstest2009 | 18.8 | 0.499 | 2525 | 62816 | 0.993 |
89
- | eng-deu | newstest2010 | 20.8 | 0.509 | 2489 | 61511 | 0.958 |
90
- | eng-deu | newstest2011 | 19.2 | 0.493 | 3003 | 72981 | 0.980 |
91
- | eng-deu | newstest2012 | 19.6 | 0.494 | 3003 | 72886 | 0.960 |
92
- | eng-deu | newstest2013 | 22.8 | 0.518 | 3000 | 63737 | 0.974 |
93
- | eng-deu | newstest2015-ende | 25.8 | 0.545 | 2169 | 44260 | 1.000 |
94
- | eng-deu | newstest2016-ende | 30.3 | 0.581 | 2999 | 62670 | 0.989 |
95
- | eng-deu | newstest2017-ende | 24.2 | 0.537 | 3004 | 61291 | 1.000 |
96
- | eng-deu | newstest2018-ende | 35.5 | 0.616 | 2998 | 64276 | 1.000 |
97
- | eng-deu | newstest2019-ende | 31.6 | 0.586 | 1997 | 48969 | 0.973 |
98
- | eng-deu | Tatoeba-test | 37.8 | 0.591 | 10000 | 83347 | 0.991 |
99
- | eng-nds | Tatoeba-test | 16.5 | 0.411 | 2500 | 18264 | 0.992 |
100
- | eng-nld | Tatoeba-test | 50.3 | 0.677 | 10000 | 71436 | 0.979 |
101
- | fry-deu | Tatoeba-test | 28.7 | 0.545 | 66 | 432 | 1.000 |
102
- | fry-eng | Tatoeba-test | 31.9 | 0.496 | 205 | 1500 | 1.000 |
103
- | fry-nld | Tatoeba-test | 43.0 | 0.634 | 233 | 1672 | 1.000 |
104
- | gos-nld | Tatoeba-test | 15.9 | 0.409 | 1852 | 9903 | 0.959 |
105
- | hrx-deu | Tatoeba-test | 24.7 | 0.487 | 471 | 2805 | 0.984 |
106
- | ltz-deu | Tatoeba-test | 36.6 | 0.552 | 337 | 2144 | 1.000 |
107
- | ltz-eng | Tatoeba-test | 31.4 | 0.477 | 283 | 1751 | 1.000 |
108
- | ltz-nld | Tatoeba-test | 37.5 | 0.523 | 273 | 1567 | 1.000 |
109
- | multi-multi | Tatoeba-test | 37.1 | 0.569 | 10000 | 73153 | 1.000 |
110
- | nds-deu | Tatoeba-test | 34.5 | 0.572 | 10000 | 74571 | 1.000 |
111
- | nds-eng | Tatoeba-test | 29.6 | 0.492 | 2500 | 17589 | 1.000 |
112
- | nds-nld | Tatoeba-test | 42.2 | 0.621 | 1657 | 11490 | 0.994 |
113
- | nld-afr | Tatoeba-test | 59.0 | 0.756 | 1056 | 6823 | 1.000 |
114
- | nld-deu | Tatoeba-test | 50.6 | 0.688 | 10000 | 72438 | 1.000 |
115
- | nld-eng | Tatoeba-test | 54.5 | 0.702 | 10000 | 69848 | 0.975 |
116
- | nld-fry | Tatoeba-test | 23.3 | 0.462 | 233 | 1679 | 1.000 |
117
- | nld-nds | Tatoeba-test | 21.7 | 0.462 | 1657 | 11711 | 0.998 |
118
- | pdc-eng | Tatoeba-test | 24.3 | 0.402 | 53 | 399 | 1.000 |
119
- | yid-nld | Tatoeba-test | 21.3 | 0.402 | 55 | 323 | 1.000 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
120
 
121
  ## Acknowledgements
122
 
@@ -125,6 +1718,6 @@ The work is supported by the [European Language Grid](https://www.european-langu
125
  ## Model conversion info
126
 
127
  * transformers version: 4.12.3
128
- * OPUS-MT git hash: b250e2e
129
- * port time: Thu Jan 27 23:08:33 EET 2022
130
  * port machine: LM0-400-22516.local
 
1
  ---
2
  language:
3
+ - af
4
+ - de
5
+ - en
6
+ - fy
7
  - gmw
8
+ - gos
9
+ - hrx
10
+ - lb
11
+ - nds
12
+ - nl
13
+ - pdc
14
+ - yi
15
+
16
  tags:
17
  - translation
18
+
19
  license: cc-by-4.0
20
  model-index:
21
  - name: opus-mt-tc-base-gmw-gmw
22
  results:
23
+ - task::
24
+ name: Translation deu-eng
25
+ type: translation
26
+ args: deu-eng
27
+ dataset:
28
+ name: WMT-news-test2008
29
+ type: WMT-news-test2008
30
+ args: deu-eng
31
+ metrics:
32
+ - name: BLEU
33
+ type: bleu
34
+ value: 23.8
35
+ - name: chr-F2
36
+ type: chrf
37
+ value: 0.518
38
+ - task::
39
+ name: Translation eng-deu
40
+ type: translation
41
+ args: eng-deu
42
+ dataset:
43
+ name: WMT-news-test2008
44
+ type: WMT-news-test2008
45
+ args: eng-deu
46
+ metrics:
47
+ - name: BLEU
48
+ type: bleu
49
+ value: 19.3
50
+ - name: chr-F2
51
+ type: chrf
52
+ value: 0.492
53
+ - task::
54
+ name: Translation deu-eng
55
+ type: translation
56
+ args: deu-eng
57
+ dataset:
58
+ name: WMT-newssyscomb2009
59
+ type: WMT-newssyscomb2009
60
+ args: deu-eng
61
+ metrics:
62
+ - name: BLEU
63
+ type: bleu
64
+ value: 25.4
65
+ - name: chr-F2
66
+ type: chrf
67
+ value: 0.527
68
+ - task::
69
+ name: Translation eng-deu
70
+ type: translation
71
+ args: eng-deu
72
+ dataset:
73
+ name: WMT-newssyscomb2009
74
+ type: WMT-newssyscomb2009
75
+ args: eng-deu
76
+ metrics:
77
+ - name: BLEU
78
+ type: bleu
79
+ value: 19.3
80
+ - name: chr-F2
81
+ type: chrf
82
+ value: 0.504
83
+ - task::
84
+ name: Translation deu-eng
85
+ type: translation
86
+ args: deu-eng
87
+ dataset:
88
+ name: WMT-newstest2009
89
+ type: WMT-newstest2009
90
+ args: deu-eng
91
+ metrics:
92
+ - name: BLEU
93
+ type: bleu
94
+ value: 23.4
95
+ - name: chr-F2
96
+ type: chrf
97
+ value: 0.516
98
+ - task::
99
+ name: Translation eng-deu
100
+ type: translation
101
+ args: eng-deu
102
+ dataset:
103
+ name: WMT-newstest2009
104
+ type: WMT-newstest2009
105
+ args: eng-deu
106
+ metrics:
107
+ - name: BLEU
108
+ type: bleu
109
+ value: 18.8
110
+ - name: chr-F2
111
+ type: chrf
112
+ value: 0.498
113
+ - task::
114
+ name: Translation deu-eng
115
+ type: translation
116
+ args: deu-eng
117
+ dataset:
118
+ name: WMT-newstest2010
119
+ type: WMT-newstest2010
120
+ args: deu-eng
121
+ metrics:
122
+ - name: BLEU
123
+ type: bleu
124
+ value: 25.8
125
+ - name: chr-F2
126
+ type: chrf
127
+ value: 0.546
128
+ - task::
129
+ name: Translation eng-deu
130
+ type: translation
131
+ args: eng-deu
132
+ dataset:
133
+ name: WMT-newstest2010
134
+ type: WMT-newstest2010
135
+ args: eng-deu
136
+ metrics:
137
+ - name: BLEU
138
+ type: bleu
139
+ value: 20.7
140
+ - name: chr-F2
141
+ type: chrf
142
+ value: 0.508
143
+ - task::
144
+ name: Translation deu-eng
145
+ type: translation
146
+ args: deu-eng
147
+ dataset:
148
+ name: WMT-newstest2011
149
+ type: WMT-newstest2011
150
+ args: deu-eng
151
+ metrics:
152
+ - name: BLEU
153
+ type: bleu
154
+ value: 23.7
155
+ - name: chr-F2
156
+ type: chrf
157
+ value: 0.524
158
+ - task::
159
+ name: Translation eng-deu
160
+ type: translation
161
+ args: eng-deu
162
+ dataset:
163
+ name: WMT-newstest2011
164
+ type: WMT-newstest2011
165
+ args: eng-deu
166
+ metrics:
167
+ - name: BLEU
168
+ type: bleu
169
+ value: 19.2
170
+ - name: chr-F2
171
+ type: chrf
172
+ value: 0.493
173
+ - task::
174
+ name: Translation deu-eng
175
+ type: translation
176
+ args: deu-eng
177
+ dataset:
178
+ name: WMT-newstest2012
179
+ type: WMT-newstest2012
180
+ args: deu-eng
181
+ metrics:
182
+ - name: BLEU
183
+ type: bleu
184
+ value: 24.8
185
+ - name: chr-F2
186
+ type: chrf
187
+ value: 0.532
188
+ - task::
189
+ name: Translation eng-deu
190
+ type: translation
191
+ args: eng-deu
192
+ dataset:
193
+ name: WMT-newstest2012
194
+ type: WMT-newstest2012
195
+ args: eng-deu
196
+ metrics:
197
+ - name: BLEU
198
+ type: bleu
199
+ value: 19.5
200
+ - name: chr-F2
201
+ type: chrf
202
+ value: 0.493
203
+ - task::
204
+ name: Translation deu-eng
205
+ type: translation
206
+ args: deu-eng
207
+ dataset:
208
+ name: WMT-newstest2013
209
+ type: WMT-newstest2013
210
+ args: deu-eng
211
+ metrics:
212
+ - name: BLEU
213
+ type: bleu
214
+ value: 27.7
215
+ - name: chr-F2
216
+ type: chrf
217
+ value: 0.548
218
+ - task::
219
+ name: Translation eng-deu
220
+ type: translation
221
+ args: eng-deu
222
+ dataset:
223
+ name: WMT-newstest2013
224
+ type: WMT-newstest2013
225
+ args: eng-deu
226
+ metrics:
227
+ - name: BLEU
228
+ type: bleu
229
+ value: 22.5
230
+ - name: chr-F2
231
+ type: chrf
232
+ value: 0.517
233
+ - task::
234
+ name: Translation deu-eng
235
+ type: translation
236
+ args: deu-eng
237
+ dataset:
238
+ name: WMT-newstest2014-deen
239
+ type: WMT-newstest2014-deen
240
+ args: deu-eng
241
+ metrics:
242
+ - name: BLEU
243
+ type: bleu
244
+ value: 27.3
245
+ - name: chr-F2
246
+ type: chrf
247
+ value: 0.548
248
+ - task::
249
+ name: Translation eng-deu
250
+ type: translation
251
+ args: eng-deu
252
+ dataset:
253
+ name: WMT-newstest2014-deen
254
+ type: WMT-newstest2014-deen
255
+ args: eng-deu
256
+ metrics:
257
+ - name: BLEU
258
+ type: bleu
259
+ value: 22.0
260
+ - name: chr-F2
261
+ type: chrf
262
+ value: 0.532
263
+ - task::
264
+ name: Translation deu-eng
265
+ type: translation
266
+ args: deu-eng
267
+ dataset:
268
+ name: WMT-newstest2015-deen
269
+ type: WMT-newstest2015-deen
270
+ args: deu-eng
271
+ metrics:
272
+ - name: BLEU
273
+ type: bleu
274
+ value: 28.6
275
+ - name: chr-F2
276
+ type: chrf
277
+ value: 0.553
278
+ - task::
279
+ name: Translation eng-deu
280
+ type: translation
281
+ args: eng-deu
282
+ dataset:
283
+ name: WMT-newstest2015-ende
284
+ type: WMT-newstest2015-ende
285
+ args: eng-deu
286
+ metrics:
287
+ - name: BLEU
288
+ type: bleu
289
+ value: 25.7
290
+ - name: chr-F2
291
+ type: chrf
292
+ value: 0.544
293
+ - task::
294
+ name: Translation deu-eng
295
+ type: translation
296
+ args: deu-eng
297
+ dataset:
298
+ name: WMT-newstest2016-deen
299
+ type: WMT-newstest2016-deen
300
+ args: deu-eng
301
+ metrics:
302
+ - name: BLEU
303
+ type: bleu
304
+ value: 33.3
305
+ - name: chr-F2
306
+ type: chrf
307
+ value: 0.596
308
+ - task::
309
+ name: Translation eng-deu
310
+ type: translation
311
+ args: eng-deu
312
+ dataset:
313
+ name: WMT-newstest2016-ende
314
+ type: WMT-newstest2016-ende
315
+ args: eng-deu
316
+ metrics:
317
+ - name: BLEU
318
+ type: bleu
319
+ value: 30.0
320
+ - name: chr-F2
321
+ type: chrf
322
+ value: 0.580
323
+ - task::
324
+ name: Translation deu-eng
325
+ type: translation
326
+ args: deu-eng
327
+ dataset:
328
+ name: WMT-newstest2017-deen
329
+ type: WMT-newstest2017-deen
330
+ args: deu-eng
331
+ metrics:
332
+ - name: BLEU
333
+ type: bleu
334
+ value: 29.5
335
+ - name: chr-F2
336
+ type: chrf
337
+ value: 0.561
338
+ - task::
339
+ name: Translation eng-deu
340
+ type: translation
341
+ args: eng-deu
342
+ dataset:
343
+ name: WMT-newstest2017-ende
344
+ type: WMT-newstest2017-ende
345
+ args: eng-deu
346
+ metrics:
347
+ - name: BLEU
348
+ type: bleu
349
+ value: 24.1
350
+ - name: chr-F2
351
+ type: chrf
352
+ value: 0.535
353
+ - task::
354
+ name: Translation deu-eng
355
+ type: translation
356
+ args: deu-eng
357
+ dataset:
358
+ name: WMT-newstest2018-deen
359
+ type: WMT-newstest2018-deen
360
+ args: deu-eng
361
+ metrics:
362
+ - name: BLEU
363
+ type: bleu
364
+ value: 36.1
365
+ - name: chr-F2
366
+ type: chrf
367
+ value: 0.610
368
+ - task::
369
+ name: Translation eng-deu
370
+ type: translation
371
+ args: eng-deu
372
+ dataset:
373
+ name: WMT-newstest2018-ende
374
+ type: WMT-newstest2018-ende
375
+ args: eng-deu
376
+ metrics:
377
+ - name: BLEU
378
+ type: bleu
379
+ value: 35.4
380
+ - name: chr-F2
381
+ type: chrf
382
+ value: 0.613
383
+ - task::
384
+ name: Translation deu-eng
385
+ type: translation
386
+ args: deu-eng
387
+ dataset:
388
+ name: WMT-newstest2019-deen
389
+ type: WMT-newstest2019-deen
390
+ args: deu-eng
391
+ metrics:
392
+ - name: BLEU
393
+ type: bleu
394
+ value: 32.3
395
+ - name: chr-F2
396
+ type: chrf
397
+ value: 0.582
398
+ - task::
399
+ name: Translation eng-deu
400
+ type: translation
401
+ args: eng-deu
402
+ dataset:
403
+ name: WMT-newstest2019-ende
404
+ type: WMT-newstest2019-ende
405
+ args: eng-deu
406
+ metrics:
407
+ - name: BLEU
408
+ type: bleu
409
+ value: 31.2
410
+ - name: chr-F2
411
+ type: chrf
412
+ value: 0.583
413
+ - task::
414
+ name: Translation deu-eng
415
+ type: translation
416
+ args: deu-eng
417
+ dataset:
418
+ name: WMT-newstest2020-deen
419
+ type: WMT-newstest2020-deen
420
+ args: deu-eng
421
+ metrics:
422
+ - name: BLEU
423
+ type: bleu
424
+ value: 32.0
425
+ - name: chr-F2
426
+ type: chrf
427
+ value: 0.604
428
+ - task::
429
+ name: Translation eng-deu
430
+ type: translation
431
+ args: eng-deu
432
+ dataset:
433
+ name: WMT-newstest2020-ende
434
+ type: WMT-newstest2020-ende
435
+ args: eng-deu
436
+ metrics:
437
+ - name: BLEU
438
+ type: bleu
439
+ value: 23.9
440
+ - name: chr-F2
441
+ type: chrf
442
+ value: 0.542
443
+ - task::
444
+ name: Translation deu-eng
445
+ type: translation
446
+ args: deu-eng
447
+ dataset:
448
+ name: WMT-newstestB2020-deen
449
+ type: WMT-newstestB2020-deen
450
+ args: deu-eng
451
+ metrics:
452
+ - name: BLEU
453
+ type: bleu
454
+ value: 31.2
455
+ - name: chr-F2
456
+ type: chrf
457
+ value: 0.598
458
+ - task::
459
+ name: Translation eng-deu
460
+ type: translation
461
+ args: eng-deu
462
+ dataset:
463
+ name: WMT-newstestB2020-ende
464
+ type: WMT-newstestB2020-ende
465
+ args: eng-deu
466
+ metrics:
467
+ - name: BLEU
468
+ type: bleu
469
+ value: 23.3
470
+ - name: chr-F2
471
+ type: chrf
472
+ value: 0.532
473
+ - task::
474
+ name: Translation afr-deu
475
+ type: translation
476
+ args: afr-deu
477
+ dataset:
478
+ name: flores101-devtest
479
+ type: flores101-devtest
480
+ args: afr-deu
481
+ metrics:
482
+ - name: BLEU
483
+ type: bleu
484
+ value: 21.6
485
+ - name: chr-F2
486
+ type: chrf
487
+ value: 0.524
488
+ - task::
489
+ name: Translation afr-eng
490
+ type: translation
491
+ args: afr-eng
492
+ dataset:
493
+ name: flores101-devtest
494
+ type: flores101-devtest
495
+ args: afr-eng
496
+ metrics:
497
+ - name: BLEU
498
+ type: bleu
499
+ value: 46.8
500
+ - name: chr-F2
501
+ type: chrf
502
+ value: 0.693
503
+ - task::
504
+ name: Translation afr-ltz
505
+ type: translation
506
+ args: afr-ltz
507
+ dataset:
508
+ name: flores101-devtest
509
+ type: flores101-devtest
510
+ args: afr-ltz
511
+ metrics:
512
+ - name: BLEU
513
+ type: bleu
514
+ value: 5.6
515
+ - name: chr-F2
516
+ type: chrf
517
+ value: 0.338
518
+ - task::
519
+ name: Translation afr-nld
520
+ type: translation
521
+ args: afr-nld
522
+ dataset:
523
+ name: flores101-devtest
524
+ type: flores101-devtest
525
+ args: afr-nld
526
+ metrics:
527
+ - name: BLEU
528
+ type: bleu
529
+ value: 18.4
530
+ - name: chr-F2
531
+ type: chrf
532
+ value: 0.509
533
+ - task::
534
+ name: Translation deu-afr
535
+ type: translation
536
+ args: deu-afr
537
+ dataset:
538
+ name: flores101-devtest
539
+ type: flores101-devtest
540
+ args: deu-afr
541
+ metrics:
542
+ - name: BLEU
543
+ type: bleu
544
+ value: 21.4
545
+ - name: chr-F2
546
+ type: chrf
547
+ value: 0.534
548
+ - task::
549
+ name: Translation deu-eng
550
+ type: translation
551
+ args: deu-eng
552
+ dataset:
553
+ name: flores101-devtest
554
+ type: flores101-devtest
555
+ args: deu-eng
556
+ metrics:
557
+ - name: BLEU
558
+ type: bleu
559
+ value: 33.8
560
+ - name: chr-F2
561
+ type: chrf
562
+ value: 0.616
563
+ - task::
564
+ name: Translation deu-ltz
565
+ type: translation
566
+ args: deu-ltz
567
+ dataset:
568
+ name: flores101-devtest
569
+ type: flores101-devtest
570
+ args: deu-ltz
571
+ metrics:
572
+ - name: BLEU
573
+ type: bleu
574
+ value: 7.9
575
+ - name: chr-F2
576
+ type: chrf
577
+ value: 0.381
578
+ - task::
579
+ name: Translation deu-nld
580
+ type: translation
581
+ args: deu-nld
582
+ dataset:
583
+ name: flores101-devtest
584
+ type: flores101-devtest
585
+ args: deu-nld
586
+ metrics:
587
+ - name: BLEU
588
+ type: bleu
589
+ value: 19.2
590
+ - name: chr-F2
591
+ type: chrf
592
+ value: 0.516
593
+ - task::
594
+ name: Translation eng-afr
595
+ type: translation
596
+ args: eng-afr
597
+ dataset:
598
+ name: flores101-devtest
599
+ type: flores101-devtest
600
+ args: eng-afr
601
+ metrics:
602
+ - name: BLEU
603
+ type: bleu
604
+ value: 33.8
605
+ - name: chr-F2
606
+ type: chrf
607
+ value: 0.628
608
+ - task::
609
+ name: Translation eng-deu
610
+ type: translation
611
+ args: eng-deu
612
+ dataset:
613
+ name: flores101-devtest
614
+ type: flores101-devtest
615
+ args: eng-deu
616
+ metrics:
617
+ - name: BLEU
618
+ type: bleu
619
+ value: 29.1
620
+ - name: chr-F2
621
+ type: chrf
622
+ value: 0.581
623
+ - task::
624
+ name: Translation eng-ltz
625
+ type: translation
626
+ args: eng-ltz
627
+ dataset:
628
+ name: flores101-devtest
629
+ type: flores101-devtest
630
+ args: eng-ltz
631
+ metrics:
632
+ - name: BLEU
633
+ type: bleu
634
+ value: 6.3
635
+ - name: chr-F2
636
+ type: chrf
637
+ value: 0.351
638
+ - task::
639
+ name: Translation eng-nld
640
+ type: translation
641
+ args: eng-nld
642
+ dataset:
643
+ name: flores101-devtest
644
+ type: flores101-devtest
645
+ args: eng-nld
646
+ metrics:
647
+ - name: BLEU
648
+ type: bleu
649
+ value: 21.0
650
+ - name: chr-F2
651
+ type: chrf
652
+ value: 0.533
653
+ - task::
654
+ name: Translation ltz-afr
655
+ type: translation
656
+ args: ltz-afr
657
+ dataset:
658
+ name: flores101-devtest
659
+ type: flores101-devtest
660
+ args: ltz-afr
661
+ metrics:
662
+ - name: BLEU
663
+ type: bleu
664
+ value: 12.9
665
+ - name: chr-F2
666
+ type: chrf
667
+ value: 0.430
668
+ - task::
669
+ name: Translation ltz-deu
670
+ type: translation
671
+ args: ltz-deu
672
+ dataset:
673
+ name: flores101-devtest
674
+ type: flores101-devtest
675
+ args: ltz-deu
676
+ metrics:
677
+ - name: BLEU
678
+ type: bleu
679
+ value: 17.1
680
+ - name: chr-F2
681
+ type: chrf
682
+ value: 0.482
683
+ - task::
684
+ name: Translation ltz-eng
685
+ type: translation
686
+ args: ltz-eng
687
+ dataset:
688
+ name: flores101-devtest
689
+ type: flores101-devtest
690
+ args: ltz-eng
691
+ metrics:
692
+ - name: BLEU
693
+ type: bleu
694
+ value: 18.8
695
+ - name: chr-F2
696
+ type: chrf
697
+ value: 0.468
698
+ - task::
699
+ name: Translation ltz-nld
700
+ type: translation
701
+ args: ltz-nld
702
+ dataset:
703
+ name: flores101-devtest
704
+ type: flores101-devtest
705
+ args: ltz-nld
706
+ metrics:
707
+ - name: BLEU
708
+ type: bleu
709
+ value: 10.7
710
+ - name: chr-F2
711
+ type: chrf
712
+ value: 0.409
713
+ - task::
714
+ name: Translation nld-afr
715
+ type: translation
716
+ args: nld-afr
717
+ dataset:
718
+ name: flores101-devtest
719
+ type: flores101-devtest
720
+ args: nld-afr
721
+ metrics:
722
+ - name: BLEU
723
+ type: bleu
724
+ value: 16.8
725
+ - name: chr-F2
726
+ type: chrf
727
+ value: 0.494
728
+ - task::
729
+ name: Translation nld-deu
730
+ type: translation
731
+ args: nld-deu
732
+ dataset:
733
+ name: flores101-devtest
734
+ type: flores101-devtest
735
+ args: nld-deu
736
+ metrics:
737
+ - name: BLEU
738
+ type: bleu
739
+ value: 17.9
740
+ - name: chr-F2
741
+ type: chrf
742
+ value: 0.501
743
+ - task::
744
+ name: Translation nld-eng
745
+ type: translation
746
+ args: nld-eng
747
+ dataset:
748
+ name: flores101-devtest
749
+ type: flores101-devtest
750
+ args: nld-eng
751
+ metrics:
752
+ - name: BLEU
753
+ type: bleu
754
+ value: 25.6
755
+ - name: chr-F2
756
+ type: chrf
757
+ value: 0.551
758
+ - task::
759
+ name: Translation nld-ltz
760
+ type: translation
761
+ args: nld-ltz
762
+ dataset:
763
+ name: flores101-devtest
764
+ type: flores101-devtest
765
+ args: nld-ltz
766
+ metrics:
767
+ - name: BLEU
768
+ type: bleu
769
+ value: 4.7
770
+ - name: chr-F2
771
+ type: chrf
772
+ value: 0.324
773
+ - task::
774
+ name: Translation deu-eng
775
+ type: translation
776
+ args: deu-eng
777
+ dataset:
778
+ name: multi30k_task2_test_2016
779
+ type: multi30k_task2_test_2016
780
+ args: deu-eng
781
+ metrics:
782
+ - name: BLEU
783
+ type: bleu
784
+ value: 3.1
785
+ - name: chr-F2
786
+ type: chrf
787
+ value: 0.201
788
+ - task::
789
+ name: Translation eng-deu
790
+ type: translation
791
+ args: eng-deu
792
+ dataset:
793
+ name: multi30k_task2_test_2016
794
+ type: multi30k_task2_test_2016
795
+ args: eng-deu
796
+ metrics:
797
+ - name: BLEU
798
+ type: bleu
799
+ value: 2.3
800
+ - name: chr-F2
801
+ type: chrf
802
+ value: 0.258
803
+ - task::
804
+ name: Translation deu-eng
805
+ type: translation
806
+ args: deu-eng
807
+ dataset:
808
+ name: multi30k_test_2016_flickr
809
+ type: multi30k_test_2016_flickr
810
+ args: deu-eng
811
+ metrics:
812
+ - name: BLEU
813
+ type: bleu
814
+ value: 32.2
815
+ - name: chr-F2
816
+ type: chrf
817
+ value: 0.546
818
+ - task::
819
+ name: Translation eng-deu
820
+ type: translation
821
+ args: eng-deu
822
+ dataset:
823
+ name: multi30k_test_2016_flickr
824
+ type: multi30k_test_2016_flickr
825
+ args: eng-deu
826
+ metrics:
827
+ - name: BLEU
828
+ type: bleu
829
+ value: 28.8
830
+ - name: chr-F2
831
+ type: chrf
832
+ value: 0.582
833
+ - task::
834
+ name: Translation deu-eng
835
+ type: translation
836
+ args: deu-eng
837
+ dataset:
838
+ name: multi30k_test_2017_flickr
839
+ type: multi30k_test_2017_flickr
840
+ args: deu-eng
841
+ metrics:
842
+ - name: BLEU
843
+ type: bleu
844
+ value: 32.7
845
+ - name: chr-F2
846
+ type: chrf
847
+ value: 0.561
848
+ - task::
849
+ name: Translation eng-deu
850
+ type: translation
851
+ args: eng-deu
852
+ dataset:
853
+ name: multi30k_test_2017_flickr
854
+ type: multi30k_test_2017_flickr
855
+ args: eng-deu
856
+ metrics:
857
+ - name: BLEU
858
+ type: bleu
859
+ value: 27.6
860
+ - name: chr-F2
861
+ type: chrf
862
+ value: 0.573
863
+ - task::
864
+ name: Translation deu-eng
865
+ type: translation
866
+ args: deu-eng
867
+ dataset:
868
+ name: multi30k_test_2017_mscoco
869
+ type: multi30k_test_2017_mscoco
870
+ args: deu-eng
871
+ metrics:
872
+ - name: BLEU
873
+ type: bleu
874
+ value: 25.5
875
+ - name: chr-F2
876
+ type: chrf
877
+ value: 0.499
878
+ - task::
879
+ name: Translation eng-deu
880
+ type: translation
881
+ args: eng-deu
882
+ dataset:
883
+ name: multi30k_test_2017_mscoco
884
+ type: multi30k_test_2017_mscoco
885
+ args: eng-deu
886
+ metrics:
887
+ - name: BLEU
888
+ type: bleu
889
+ value: 22.0
890
+ - name: chr-F2
891
+ type: chrf
892
+ value: 0.514
893
+ - task::
894
+ name: Translation deu-eng
895
+ type: translation
896
+ args: deu-eng
897
+ dataset:
898
+ name: multi30k_test_2018_flickr
899
+ type: multi30k_test_2018_flickr
900
+ args: deu-eng
901
+ metrics:
902
+ - name: BLEU
903
+ type: bleu
904
+ value: 30.0
905
+ - name: chr-F2
906
+ type: chrf
907
+ value: 0.535
908
+ - task::
909
+ name: Translation eng-deu
910
+ type: translation
911
+ args: eng-deu
912
+ dataset:
913
+ name: multi30k_test_2018_flickr
914
+ type: multi30k_test_2018_flickr
915
+ args: eng-deu
916
+ metrics:
917
+ - name: BLEU
918
+ type: bleu
919
+ value: 25.3
920
+ - name: chr-F2
921
+ type: chrf
922
+ value: 0.547
923
+ - task::
924
+ name: Translation afr-deu
925
  type: translation
926
  args: afr-deu
927
  dataset:
928
+ name: tatoeba-test
929
+ type: tatoeba-test
930
+ args: afr-deu v2021-08-07
931
+ metrics:
932
+ - name: BLEU
933
+ type: bleu
934
+ value: 48.1
935
+ - name: chr-F2
936
+ type: chrf
937
+ value: 0.674
938
+ - task::
939
+ name: Translation afr-eng
940
+ type: translation
941
+ args: afr-eng
942
+ dataset:
943
+ name: tatoeba-test
944
+ type: tatoeba-test
945
+ args: afr-eng v2021-08-07
946
+ metrics:
947
+ - name: BLEU
948
+ type: bleu
949
+ value: 58.8
950
+ - name: chr-F2
951
+ type: chrf
952
+ value: 0.728
953
+ - task::
954
+ name: Translation afr-nld
955
+ type: translation
956
+ args: afr-nld
957
+ dataset:
958
+ name: tatoeba-test
959
+ type: tatoeba-test
960
+ args: afr-nld v2021-08-07
961
+ metrics:
962
+ - name: BLEU
963
+ type: bleu
964
+ value: 54.5
965
+ - name: chr-F2
966
+ type: chrf
967
+ value: 0.711
968
+ - task::
969
+ name: Translation ang-eng
970
+ type: translation
971
+ args: ang-eng
972
+ dataset:
973
+ name: tatoeba-test
974
+ type: tatoeba-test
975
+ args: ang-eng v2021-03-30
976
+ metrics:
977
+ - name: BLEU
978
+ type: bleu
979
+ value: 7.8
980
+ - name: chr-F2
981
+ type: chrf
982
+ value: 0.215
983
+ - task::
984
+ name: Translation deu-afr
985
+ type: translation
986
+ args: deu-afr
987
+ dataset:
988
+ name: tatoeba-test
989
+ type: tatoeba-test
990
+ args: deu-afr v2021-08-07
991
+ metrics:
992
+ - name: BLEU
993
+ type: bleu
994
+ value: 52.4
995
+ - name: chr-F2
996
+ type: chrf
997
+ value: 0.696
998
+ - task::
999
+ name: Translation deu-eng
1000
+ type: translation
1001
+ args: deu-eng
1002
+ dataset:
1003
+ name: tatoeba-test
1004
+ type: tatoeba-test
1005
+ args: deu-eng v2021-08-07
1006
+ metrics:
1007
+ - name: BLEU
1008
+ type: bleu
1009
+ value: 42.1
1010
+ - name: chr-F2
1011
+ type: chrf
1012
+ value: 0.609
1013
+ - task::
1014
+ name: Translation deu-ltz
1015
+ type: translation
1016
+ args: deu-ltz
1017
+ dataset:
1018
+ name: tatoeba-test
1019
+ type: tatoeba-test
1020
+ args: deu-ltz v2021-08-07
1021
+ metrics:
1022
+ - name: BLEU
1023
+ type: bleu
1024
+ value: 13.4
1025
+ - name: chr-F2
1026
+ type: chrf
1027
+ value: 0.282
1028
+ - task::
1029
+ name: Translation deu-nds
1030
+ type: translation
1031
+ args: deu-nds
1032
+ dataset:
1033
+ name: tatoeba-test
1034
+ type: tatoeba-test
1035
+ args: deu-nds v2021-08-07
1036
+ metrics:
1037
+ - name: BLEU
1038
+ type: bleu
1039
+ value: 18.6
1040
+ - name: chr-F2
1041
+ type: chrf
1042
+ value: 0.442
1043
+ - task::
1044
+ name: Translation deu-nld
1045
+ type: translation
1046
+ args: deu-nld
1047
+ dataset:
1048
+ name: tatoeba-test
1049
+ type: tatoeba-test
1050
+ args: deu-nld v2021-08-07
1051
+ metrics:
1052
+ - name: BLEU
1053
+ type: bleu
1054
+ value: 48.7
1055
+ - name: chr-F2
1056
+ type: chrf
1057
+ value: 0.672
1058
+ - task::
1059
+ name: Translation deu-yid
1060
+ type: translation
1061
+ args: deu-yid
1062
+ dataset:
1063
+ name: tatoeba-test
1064
+ type: tatoeba-test
1065
+ args: deu-yid v2021-08-07
1066
+ metrics:
1067
+ - name: BLEU
1068
+ type: bleu
1069
+ value: 2.3
1070
+ - name: chr-F2
1071
+ type: chrf
1072
+ value: 0.198
1073
+ - task::
1074
+ name: Translation eng-afr
1075
+ type: translation
1076
+ args: eng-afr
1077
+ dataset:
1078
+ name: tatoeba-test
1079
+ type: tatoeba-test
1080
+ args: eng-afr v2021-08-07
1081
+ metrics:
1082
+ - name: BLEU
1083
+ type: bleu
1084
+ value: 56.5
1085
+ - name: chr-F2
1086
+ type: chrf
1087
+ value: 0.735
1088
+ - task::
1089
+ name: Translation eng-deu
1090
+ type: translation
1091
+ args: eng-deu
1092
+ dataset:
1093
+ name: tatoeba-test
1094
+ type: tatoeba-test
1095
+ args: eng-deu v2021-08-07
1096
+ metrics:
1097
+ - name: BLEU
1098
+ type: bleu
1099
+ value: 35.9
1100
+ - name: chr-F2
1101
+ type: chrf
1102
+ value: 0.580
1103
+ - task::
1104
+ name: Translation eng-fry
1105
+ type: translation
1106
+ args: eng-fry
1107
+ dataset:
1108
+ name: tatoeba-test
1109
+ type: tatoeba-test
1110
+ args: eng-fry v2021-08-07
1111
+ metrics:
1112
+ - name: BLEU
1113
+ type: bleu
1114
+ value: 17.2
1115
+ - name: chr-F2
1116
+ type: chrf
1117
+ value: 0.389
1118
+ - task::
1119
+ name: Translation eng-ltz
1120
+ type: translation
1121
+ args: eng-ltz
1122
+ dataset:
1123
+ name: tatoeba-test
1124
+ type: tatoeba-test
1125
+ args: eng-ltz v2021-08-07
1126
+ metrics:
1127
+ - name: BLEU
1128
+ type: bleu
1129
+ value: 14.0
1130
+ - name: chr-F2
1131
+ type: chrf
1132
+ value: 0.288
1133
+ - task::
1134
+ name: Translation eng-nds
1135
+ type: translation
1136
+ args: eng-nds
1137
+ dataset:
1138
+ name: tatoeba-test
1139
+ type: tatoeba-test
1140
+ args: eng-nds v2021-08-07
1141
+ metrics:
1142
+ - name: BLEU
1143
+ type: bleu
1144
+ value: 16.6
1145
+ - name: chr-F2
1146
+ type: chrf
1147
+ value: 0.412
1148
+ - task::
1149
+ name: Translation eng-nld
1150
+ type: translation
1151
+ args: eng-nld
1152
+ dataset:
1153
+ name: tatoeba-test
1154
+ type: tatoeba-test
1155
+ args: eng-nld v2021-08-07
1156
+ metrics:
1157
+ - name: BLEU
1158
+ type: bleu
1159
+ value: 48.3
1160
+ - name: chr-F2
1161
+ type: chrf
1162
+ value: 0.663
1163
+ - task::
1164
+ name: Translation eng-yid
1165
+ type: translation
1166
+ args: eng-yid
1167
+ dataset:
1168
+ name: tatoeba-test
1169
+ type: tatoeba-test
1170
+ args: eng-yid v2021-08-07
1171
+ metrics:
1172
+ - name: BLEU
1173
+ type: bleu
1174
+ value: 2.8
1175
+ - name: chr-F2
1176
+ type: chrf
1177
+ value: 0.207
1178
+ - task::
1179
+ name: Translation frr-deu
1180
+ type: translation
1181
+ args: frr-deu
1182
+ dataset:
1183
+ name: tatoeba-test
1184
+ type: tatoeba-test
1185
+ args: frr-deu v2021-08-07
1186
  metrics:
1187
  - name: BLEU
1188
  type: bleu
1189
+ value: 3.0
1190
+ - name: chr-F2
1191
+ type: chrf
1192
+ value: 0.198
1193
+ - task::
1194
+ name: Translation fry-eng
1195
+ type: translation
1196
+ args: fry-eng
1197
+ dataset:
1198
+ name: tatoeba-test
1199
+ type: tatoeba-test
1200
+ args: fry-eng v2021-08-07
1201
+ metrics:
1202
+ - name: BLEU
1203
+ type: bleu
1204
+ value: 32.5
1205
+ - name: chr-F2
1206
+ type: chrf
1207
+ value: 0.500
1208
+ - task::
1209
+ name: Translation fry-nld
1210
+ type: translation
1211
+ args: fry-nld
1212
+ dataset:
1213
+ name: tatoeba-test
1214
+ type: tatoeba-test
1215
+ args: fry-nld v2021-08-07
1216
+ metrics:
1217
+ - name: BLEU
1218
+ type: bleu
1219
+ value: 43.1
1220
+ - name: chr-F2
1221
+ type: chrf
1222
+ value: 0.633
1223
+ - task::
1224
+ name: Translation gos-deu
1225
+ type: translation
1226
+ args: gos-deu
1227
+ dataset:
1228
+ name: tatoeba-test
1229
+ type: tatoeba-test
1230
+ args: gos-deu v2021-08-07
1231
+ metrics:
1232
+ - name: BLEU
1233
+ type: bleu
1234
+ value: 12.7
1235
+ - name: chr-F2
1236
+ type: chrf
1237
+ value: 0.370
1238
+ - task::
1239
+ name: Translation gos-eng
1240
+ type: translation
1241
+ args: gos-eng
1242
+ dataset:
1243
+ name: tatoeba-test
1244
+ type: tatoeba-test
1245
+ args: gos-eng v2021-08-07
1246
+ metrics:
1247
+ - name: BLEU
1248
+ type: bleu
1249
+ value: 14.6
1250
+ - name: chr-F2
1251
+ type: chrf
1252
+ value: 0.316
1253
+ - task::
1254
+ name: Translation gos-nld
1255
+ type: translation
1256
+ args: gos-nld
1257
+ dataset:
1258
+ name: tatoeba-test
1259
+ type: tatoeba-test
1260
+ args: gos-nld v2021-08-07
1261
+ metrics:
1262
+ - name: BLEU
1263
+ type: bleu
1264
+ value: 15.6
1265
+ - name: chr-F2
1266
+ type: chrf
1267
+ value: 0.405
1268
+ - task::
1269
+ name: Translation gsw-eng
1270
+ type: translation
1271
+ args: gsw-eng
1272
+ dataset:
1273
+ name: tatoeba-test
1274
+ type: tatoeba-test
1275
+ args: gsw-eng v2021-08-07
1276
+ metrics:
1277
+ - name: BLEU
1278
+ type: bleu
1279
+ value: 14.0
1280
+ - name: chr-F2
1281
+ type: chrf
1282
+ value: 0.312
1283
+ - task::
1284
+ name: Translation hrx-deu
1285
+ type: translation
1286
+ args: hrx-deu
1287
+ dataset:
1288
+ name: tatoeba-test
1289
+ type: tatoeba-test
1290
+ args: hrx-deu v2021-08-07
1291
+ metrics:
1292
+ - name: BLEU
1293
+ type: bleu
1294
+ value: 24.7
1295
+ - name: chr-F2
1296
+ type: chrf
1297
+ value: 0.484
1298
+ - task::
1299
+ name: Translation hrx-eng
1300
+ type: translation
1301
+ args: hrx-eng
1302
+ dataset:
1303
+ name: tatoeba-test
1304
+ type: tatoeba-test
1305
+ args: hrx-eng v2021-08-07
1306
+ metrics:
1307
+ - name: BLEU
1308
+ type: bleu
1309
+ value: 20.4
1310
+ - name: chr-F2
1311
+ type: chrf
1312
+ value: 0.362
1313
+ - task::
1314
+ name: Translation ltz-deu
1315
+ type: translation
1316
+ args: ltz-deu
1317
+ dataset:
1318
+ name: tatoeba-test
1319
+ type: tatoeba-test
1320
+ args: ltz-deu v2021-08-07
1321
+ metrics:
1322
+ - name: BLEU
1323
+ type: bleu
1324
+ value: 37.2
1325
+ - name: chr-F2
1326
+ type: chrf
1327
+ value: 0.556
1328
+ - task::
1329
+ name: Translation ltz-eng
1330
+ type: translation
1331
+ args: ltz-eng
1332
+ dataset:
1333
+ name: tatoeba-test
1334
+ type: tatoeba-test
1335
+ args: ltz-eng v2021-08-07
1336
+ metrics:
1337
+ - name: BLEU
1338
+ type: bleu
1339
+ value: 32.4
1340
+ - name: chr-F2
1341
+ type: chrf
1342
+ value: 0.485
1343
+ - task::
1344
+ name: Translation ltz-nld
1345
+ type: translation
1346
+ args: ltz-nld
1347
+ dataset:
1348
+ name: tatoeba-test
1349
+ type: tatoeba-test
1350
+ args: ltz-nld v2021-08-07
1351
+ metrics:
1352
+ - name: BLEU
1353
+ type: bleu
1354
+ value: 39.3
1355
+ - name: chr-F2
1356
+ type: chrf
1357
+ value: 0.534
1358
+ - task::
1359
+ name: Translation nds-deu
1360
+ type: translation
1361
+ args: nds-deu
1362
+ dataset:
1363
+ name: tatoeba-test
1364
+ type: tatoeba-test
1365
+ args: nds-deu v2021-08-07
1366
+ metrics:
1367
+ - name: BLEU
1368
+ type: bleu
1369
+ value: 34.5
1370
+ - name: chr-F2
1371
+ type: chrf
1372
+ value: 0.572
1373
+ - task::
1374
+ name: Translation nds-eng
1375
+ type: translation
1376
+ args: nds-eng
1377
+ dataset:
1378
+ name: tatoeba-test
1379
+ type: tatoeba-test
1380
+ args: nds-eng v2021-08-07
1381
+ metrics:
1382
+ - name: BLEU
1383
+ type: bleu
1384
+ value: 29.9
1385
+ - name: chr-F2
1386
+ type: chrf
1387
+ value: 0.493
1388
+ - task::
1389
+ name: Translation nds-nld
1390
+ type: translation
1391
+ args: nds-nld
1392
+ dataset:
1393
+ name: tatoeba-test
1394
+ type: tatoeba-test
1395
+ args: nds-nld v2021-08-07
1396
+ metrics:
1397
+ - name: BLEU
1398
+ type: bleu
1399
+ value: 42.3
1400
+ - name: chr-F2
1401
+ type: chrf
1402
+ value: 0.621
1403
+ - task::
1404
+ name: Translation nld-afr
1405
+ type: translation
1406
+ args: nld-afr
1407
+ dataset:
1408
+ name: tatoeba-test
1409
+ type: tatoeba-test
1410
+ args: nld-afr v2021-08-07
1411
+ metrics:
1412
+ - name: BLEU
1413
+ type: bleu
1414
+ value: 58.8
1415
+ - name: chr-F2
1416
+ type: chrf
1417
+ value: 0.755
1418
+ - task::
1419
+ name: Translation nld-deu
1420
+ type: translation
1421
+ args: nld-deu
1422
+ dataset:
1423
+ name: tatoeba-test
1424
+ type: tatoeba-test
1425
+ args: nld-deu v2021-08-07
1426
+ metrics:
1427
+ - name: BLEU
1428
+ type: bleu
1429
+ value: 50.4
1430
+ - name: chr-F2
1431
+ type: chrf
1432
+ value: 0.686
1433
+ - task::
1434
+ name: Translation nld-eng
1435
+ type: translation
1436
+ args: nld-eng
1437
+ dataset:
1438
+ name: tatoeba-test
1439
+ type: tatoeba-test
1440
+ args: nld-eng v2021-08-07
1441
+ metrics:
1442
+ - name: BLEU
1443
+ type: bleu
1444
+ value: 53.1
1445
+ - name: chr-F2
1446
+ type: chrf
1447
+ value: 0.690
1448
+ - task::
1449
+ name: Translation nld-fry
1450
+ type: translation
1451
+ args: nld-fry
1452
+ dataset:
1453
+ name: tatoeba-test
1454
+ type: tatoeba-test
1455
+ args: nld-fry v2021-08-07
1456
+ metrics:
1457
+ - name: BLEU
1458
+ type: bleu
1459
+ value: 25.1
1460
+ - name: chr-F2
1461
+ type: chrf
1462
+ value: 0.478
1463
+ - task::
1464
+ name: Translation nld-ltz
1465
+ type: translation
1466
+ args: nld-ltz
1467
+ dataset:
1468
+ name: tatoeba-test
1469
+ type: tatoeba-test
1470
+ args: nld-ltz v2021-08-07
1471
+ metrics:
1472
+ - name: BLEU
1473
+ type: bleu
1474
+ value: 14.1
1475
+ - name: chr-F2
1476
+ type: chrf
1477
+ value: 0.301
1478
+ - task::
1479
+ name: Translation nld-nds
1480
+ type: translation
1481
+ args: nld-nds
1482
+ dataset:
1483
+ name: tatoeba-test
1484
+ type: tatoeba-test
1485
+ args: nld-nds v2021-08-07
1486
+ metrics:
1487
+ - name: BLEU
1488
+ type: bleu
1489
+ value: 21.4
1490
+ - name: chr-F2
1491
+ type: chrf
1492
+ value: 0.462
1493
+ - task::
1494
+ name: Translation swg-deu
1495
+ type: translation
1496
+ args: swg-deu
1497
+ dataset:
1498
+ name: tatoeba-test
1499
+ type: tatoeba-test
1500
+ args: swg-deu v2021-08-07
1501
+ metrics:
1502
+ - name: BLEU
1503
+ type: bleu
1504
+ value: 8.6
1505
+ - name: chr-F2
1506
+ type: chrf
1507
+ value: 0.335
1508
+ - task::
1509
+ name: Translation yid-deu
1510
+ type: translation
1511
+ args: yid-deu
1512
+ dataset:
1513
+ name: tatoeba-test
1514
+ type: tatoeba-test
1515
+ args: yid-deu v2021-08-07
1516
+ metrics:
1517
+ - name: BLEU
1518
+ type: bleu
1519
+ value: 17.6
1520
+ - name: chr-F2
1521
+ type: chrf
1522
+ value: 0.381
1523
+ - task::
1524
+ name: Translation yid-eng
1525
+ type: translation
1526
+ args: yid-eng
1527
+ dataset:
1528
+ name: tatoeba-test
1529
+ type: tatoeba-test
1530
+ args: yid-eng v2021-08-07
1531
+ metrics:
1532
+ - name: BLEU
1533
+ type: bleu
1534
+ value: 14.2
1535
+ - name: chr-F2
1536
+ type: chrf
1537
+ value: 0.324
1538
  ---
1539
  # opus-mt-tc-base-gmw-gmw
1540
 
 
1544
 
1545
  * 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.)
1546
 
1547
+ ```
1548
+ @inproceedings{tiedemann-thottingal-2020-opus,
1549
+ title = "{OPUS}-{MT} {--} Building open translation services for the World",
1550
+ author = {Tiedemann, J{\"o}rg and Thottingal, Santhosh},
1551
+ booktitle = "Proceedings of the 22nd Annual Conference of the European Association for Machine Translation",
1552
+ month = nov,
1553
+ year = "2020",
1554
+ address = "Lisboa, Portugal",
1555
+ publisher = "European Association for Machine Translation",
1556
+ url = "https://aclanthology.org/2020.eamt-1.61",
1557
+ pages = "479--480",
1558
+ }
1559
+
1560
+ @inproceedings{tiedemann-2020-tatoeba,
1561
+ title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
1562
+ author = {Tiedemann, J{\"o}rg},
1563
+ booktitle = "Proceedings of the Fifth Conference on Machine Translation",
1564
+ month = nov,
1565
+ year = "2020",
1566
+ address = "Online",
1567
+ publisher = "Association for Computational Linguistics",
1568
+ url = "https://aclanthology.org/2020.wmt-1.139",
1569
+ pages = "1174--1182",
1570
+ }
1571
+ ```
1572
+
1573
  ## Model info
1574
 
1575
  * Release: 2021-02-23
 
1580
  * data: opus ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
1581
  * tokenization: SentencePiece (spm32k,spm32k)
1582
  * original model: [opus-2021-02-23.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/gmw-gmw/opus-2021-02-23.zip)
1583
+ * more information released models: [OPUS-MT gmw-gmw README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/gmw-gmw/README.md)
1584
+ * more information about the model: [MarianMT](https://huggingface.co/docs/transformers/model_doc/marian)
1585
 
1586
  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. `>>afr<<`
1587
 
1588
  ## Usage
1589
 
1590
+ A short example code:
1591
+
1592
+ ```python
1593
+ from transformers import MarianMTModel, MarianTokenizer
1594
+
1595
+ src_text = [
1596
+ ">>nld<< You need help.",
1597
+ ">>afr<< I love your son."
1598
+ ]
1599
+
1600
+ model_name = "pytorch-models/opus-mt-tc-base-en-fi"
1601
+ tokenizer = MarianTokenizer.from_pretrained(model_name)
1602
+ model = MarianMTModel.from_pretrained(model_name)
1603
+ translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))
1604
+
1605
+ for t in translated:
1606
+ print( tokenizer.decode(t, skip_special_tokens=True) )
1607
+
1608
+ # expected output:
1609
+ # Je hebt hulp nodig.
1610
+ # Ek is lief vir jou seun.
1611
+ ```
1612
+
1613
+ You can also use OPUS-MT models with the transformers pipelines, for example:
1614
 
1615
  ```python
1616
  from transformers import pipeline
1617
  pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-base-gmw-gmw")
1618
+ print(pipe(>>nld<< You need help.))
1619
+
1620
+ # expected output: Je hebt hulp nodig.
1621
  ```
1622
 
1623
  ## Benchmarks
1624
 
1625
  * test set translations: [opus-2021-02-23.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/gmw-gmw/opus-2021-02-23.test.txt)
1626
  * test set scores: [opus-2021-02-23.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/gmw-gmw/opus-2021-02-23.eval.txt)
1627
+ * benchmark results: [benchmarks.tsv](benchmarks.tsv)
1628
+ * benchmark output: [benchmark_translations.zip](benchmark_translations.zip)
1629
 
1630
+ | langpair | testset | chr-F | BLEU | #sent | #words |
1631
+ |----------|---------|-------|-------|-------|--------|
1632
+ | afr-deu | tatoeba-test-v2021-08-07 | 0.674 | 48.1 | 1583 | 9105 |
1633
+ | afr-eng | tatoeba-test-v2021-08-07 | 0.728 | 58.8 | 1374 | 9622 |
1634
+ | afr-nld | tatoeba-test-v2021-08-07 | 0.711 | 54.5 | 1056 | 6710 |
1635
+ | deu-afr | tatoeba-test-v2021-08-07 | 0.696 | 52.4 | 1583 | 9507 |
1636
+ | deu-eng | tatoeba-test-v2021-08-07 | 0.609 | 42.1 | 17565 | 149462 |
1637
+ | deu-nds | tatoeba-test-v2021-08-07 | 0.442 | 18.6 | 9999 | 76137 |
1638
+ | deu-nld | tatoeba-test-v2021-08-07 | 0.672 | 48.7 | 10218 | 75235 |
1639
+ | eng-afr | tatoeba-test-v2021-08-07 | 0.735 | 56.5 | 1374 | 10317 |
1640
+ | eng-deu | tatoeba-test-v2021-08-07 | 0.580 | 35.9 | 17565 | 151568 |
1641
+ | eng-nds | tatoeba-test-v2021-08-07 | 0.412 | 16.6 | 2500 | 18264 |
1642
+ | eng-nld | tatoeba-test-v2021-08-07 | 0.663 | 48.3 | 12696 | 91796 |
1643
+ | fry-eng | tatoeba-test-v2021-08-07 | 0.500 | 32.5 | 220 | 1573 |
1644
+ | fry-nld | tatoeba-test-v2021-08-07 | 0.633 | 43.1 | 260 | 1854 |
1645
+ | gos-nld | tatoeba-test-v2021-08-07 | 0.405 | 15.6 | 1852 | 9903 |
1646
+ | hrx-deu | tatoeba-test-v2021-08-07 | 0.484 | 24.7 | 471 | 2805 |
1647
+ | hrx-eng | tatoeba-test-v2021-08-07 | 0.362 | 20.4 | 221 | 1235 |
1648
+ | ltz-deu | tatoeba-test-v2021-08-07 | 0.556 | 37.2 | 347 | 2208 |
1649
+ | ltz-eng | tatoeba-test-v2021-08-07 | 0.485 | 32.4 | 293 | 1840 |
1650
+ | ltz-nld | tatoeba-test-v2021-08-07 | 0.534 | 39.3 | 292 | 1685 |
1651
+ | nds-deu | tatoeba-test-v2021-08-07 | 0.572 | 34.5 | 9999 | 74564 |
1652
+ | nds-eng | tatoeba-test-v2021-08-07 | 0.493 | 29.9 | 2500 | 17589 |
1653
+ | nds-nld | tatoeba-test-v2021-08-07 | 0.621 | 42.3 | 1657 | 11490 |
1654
+ | nld-afr | tatoeba-test-v2021-08-07 | 0.755 | 58.8 | 1056 | 6823 |
1655
+ | nld-deu | tatoeba-test-v2021-08-07 | 0.686 | 50.4 | 10218 | 74131 |
1656
+ | nld-eng | tatoeba-test-v2021-08-07 | 0.690 | 53.1 | 12696 | 89978 |
1657
+ | nld-fry | tatoeba-test-v2021-08-07 | 0.478 | 25.1 | 260 | 1857 |
1658
+ | nld-nds | tatoeba-test-v2021-08-07 | 0.462 | 21.4 | 1657 | 11711 |
1659
+ | afr-deu | flores101-devtest | 0.524 | 21.6 | 1012 | 25094 |
1660
+ | afr-eng | flores101-devtest | 0.693 | 46.8 | 1012 | 24721 |
1661
+ | afr-nld | flores101-devtest | 0.509 | 18.4 | 1012 | 25467 |
1662
+ | deu-afr | flores101-devtest | 0.534 | 21.4 | 1012 | 25740 |
1663
+ | deu-eng | flores101-devtest | 0.616 | 33.8 | 1012 | 24721 |
1664
+ | deu-nld | flores101-devtest | 0.516 | 19.2 | 1012 | 25467 |
1665
+ | eng-afr | flores101-devtest | 0.628 | 33.8 | 1012 | 25740 |
1666
+ | eng-deu | flores101-devtest | 0.581 | 29.1 | 1012 | 25094 |
1667
+ | eng-nld | flores101-devtest | 0.533 | 21.0 | 1012 | 25467 |
1668
+ | ltz-afr | flores101-devtest | 0.430 | 12.9 | 1012 | 25740 |
1669
+ | ltz-deu | flores101-devtest | 0.482 | 17.1 | 1012 | 25094 |
1670
+ | ltz-eng | flores101-devtest | 0.468 | 18.8 | 1012 | 24721 |
1671
+ | ltz-nld | flores101-devtest | 0.409 | 10.7 | 1012 | 25467 |
1672
+ | nld-afr | flores101-devtest | 0.494 | 16.8 | 1012 | 25740 |
1673
+ | nld-deu | flores101-devtest | 0.501 | 17.9 | 1012 | 25094 |
1674
+ | nld-eng | flores101-devtest | 0.551 | 25.6 | 1012 | 24721 |
1675
+ | deu-eng | multi30k_test_2016_flickr | 0.546 | 32.2 | 1000 | 12955 |
1676
+ | eng-deu | multi30k_test_2016_flickr | 0.582 | 28.8 | 1000 | 12106 |
1677
+ | deu-eng | multi30k_test_2017_flickr | 0.561 | 32.7 | 1000 | 11374 |
1678
+ | eng-deu | multi30k_test_2017_flickr | 0.573 | 27.6 | 1000 | 10755 |
1679
+ | deu-eng | multi30k_test_2017_mscoco | 0.499 | 25.5 | 461 | 5231 |
1680
+ | eng-deu | multi30k_test_2017_mscoco | 0.514 | 22.0 | 461 | 5158 |
1681
+ | deu-eng | multi30k_test_2018_flickr | 0.535 | 30.0 | 1071 | 14689 |
1682
+ | eng-deu | multi30k_test_2018_flickr | 0.547 | 25.3 | 1071 | 13703 |
1683
+ | deu-eng | newssyscomb2009 | 0.527 | 25.4 | 502 | 11818 |
1684
+ | eng-deu | newssyscomb2009 | 0.504 | 19.3 | 502 | 11271 |
1685
+ | deu-eng | news-test2008 | 0.518 | 23.8 | 2051 | 49380 |
1686
+ | eng-deu | news-test2008 | 0.492 | 19.3 | 2051 | 47447 |
1687
+ | deu-eng | newstest2009 | 0.516 | 23.4 | 2525 | 65399 |
1688
+ | eng-deu | newstest2009 | 0.498 | 18.8 | 2525 | 62816 |
1689
+ | deu-eng | newstest2010 | 0.546 | 25.8 | 2489 | 61711 |
1690
+ | eng-deu | newstest2010 | 0.508 | 20.7 | 2489 | 61503 |
1691
+ | deu-eng | newstest2011 | 0.524 | 23.7 | 3003 | 74681 |
1692
+ | eng-deu | newstest2011 | 0.493 | 19.2 | 3003 | 72981 |
1693
+ | deu-eng | newstest2012 | 0.532 | 24.8 | 3003 | 72812 |
1694
+ | eng-deu | newstest2012 | 0.493 | 19.5 | 3003 | 72886 |
1695
+ | deu-eng | newstest2013 | 0.548 | 27.7 | 3000 | 64505 |
1696
+ | eng-deu | newstest2013 | 0.517 | 22.5 | 3000 | 63737 |
1697
+ | deu-eng | newstest2014-deen | 0.548 | 27.3 | 3003 | 67337 |
1698
+ | eng-deu | newstest2014-deen | 0.532 | 22.0 | 3003 | 62688 |
1699
+ | deu-eng | newstest2015-deen | 0.553 | 28.6 | 2169 | 46443 |
1700
+ | eng-deu | newstest2015-ende | 0.544 | 25.7 | 2169 | 44260 |
1701
+ | deu-eng | newstest2016-deen | 0.596 | 33.3 | 2999 | 64119 |
1702
+ | eng-deu | newstest2016-ende | 0.580 | 30.0 | 2999 | 62669 |
1703
+ | deu-eng | newstest2017-deen | 0.561 | 29.5 | 3004 | 64399 |
1704
+ | eng-deu | newstest2017-ende | 0.535 | 24.1 | 3004 | 61287 |
1705
+ | deu-eng | newstest2018-deen | 0.610 | 36.1 | 2998 | 67012 |
1706
+ | eng-deu | newstest2018-ende | 0.613 | 35.4 | 2998 | 64276 |
1707
+ | deu-eng | newstest2019-deen | 0.582 | 32.3 | 2000 | 39227 |
1708
+ | eng-deu | newstest2019-ende | 0.583 | 31.2 | 1997 | 48746 |
1709
+ | deu-eng | newstest2020-deen | 0.604 | 32.0 | 785 | 38220 |
1710
+ | eng-deu | newstest2020-ende | 0.542 | 23.9 | 1418 | 52383 |
1711
+ | deu-eng | newstestB2020-deen | 0.598 | 31.2 | 785 | 37696 |
1712
+ | eng-deu | newstestB2020-ende | 0.532 | 23.3 | 1418 | 53092 |
1713
 
1714
  ## Acknowledgements
1715
 
 
1718
  ## Model conversion info
1719
 
1720
  * transformers version: 4.12.3
1721
+ * OPUS-MT git hash: 64dc362
1722
+ * port time: Fri Feb 11 00:08:20 EET 2022
1723
  * port machine: LM0-400-22516.local
benchmark_translations.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a99fdfc9a2526feb7568e5e87cefd4fb1417c9cbae893733eb24928f99d351d
3
+ size 30803803
benchmarks.tsv ADDED
@@ -0,0 +1,199 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ afr-deu flores101-dev 0.516 22.2 997 24097
2
+ afr-eng flores101-dev 0.696 47.0 997 23555
3
+ afr-ltz flores101-dev 0.337 6.2 997 23906
4
+ afr-nld flores101-dev 0.509 18.6 997 24342
5
+ deu-afr flores101-dev 0.522 21.1 997 24677
6
+ deu-eng flores101-dev 0.617 34.5 997 23555
7
+ deu-ltz flores101-dev 0.381 8.8 997 23906
8
+ deu-nld flores101-dev 0.515 20.0 997 24342
9
+ eng-afr flores101-dev 0.625 33.2 997 24677
10
+ eng-deu flores101-dev 0.575 29.2 997 24097
11
+ eng-ltz flores101-dev 0.351 6.8 997 23906
12
+ eng-nld flores101-dev 0.535 21.6 997 24342
13
+ ltz-afr flores101-dev 0.430 13.7 997 24677
14
+ ltz-deu flores101-dev 0.485 18.5 997 24097
15
+ ltz-eng flores101-dev 0.474 20.1 997 23555
16
+ ltz-nld flores101-dev 0.412 11.5 997 24342
17
+ nld-afr flores101-dev 0.493 16.8 997 24677
18
+ nld-deu flores101-dev 0.494 19.0 997 24097
19
+ nld-eng flores101-dev 0.552 25.5 997 23555
20
+ nld-ltz flores101-dev 0.324 5.5 997 23906
21
+ afr-deu flores101-devtest 0.524 21.6 1012 25094
22
+ afr-eng flores101-devtest 0.693 46.8 1012 24721
23
+ afr-ltz flores101-devtest 0.338 5.6 1012 25087
24
+ afr-nld flores101-devtest 0.509 18.4 1012 25467
25
+ deu-afr flores101-devtest 0.534 21.4 1012 25740
26
+ deu-eng flores101-devtest 0.616 33.8 1012 24721
27
+ deu-ltz flores101-devtest 0.381 7.9 1012 25087
28
+ deu-nld flores101-devtest 0.516 19.2 1012 25467
29
+ eng-afr flores101-devtest 0.628 33.8 1012 25740
30
+ eng-deu flores101-devtest 0.581 29.1 1012 25094
31
+ eng-ltz flores101-devtest 0.351 6.3 1012 25087
32
+ eng-nld flores101-devtest 0.533 21.0 1012 25467
33
+ ltz-afr flores101-devtest 0.430 12.9 1012 25740
34
+ ltz-deu flores101-devtest 0.482 17.1 1012 25094
35
+ ltz-eng flores101-devtest 0.468 18.8 1012 24721
36
+ ltz-nld flores101-devtest 0.409 10.7 1012 25467
37
+ nld-afr flores101-devtest 0.494 16.8 1012 25740
38
+ nld-deu flores101-devtest 0.501 17.9 1012 25094
39
+ nld-eng flores101-devtest 0.551 25.6 1012 24721
40
+ nld-ltz flores101-devtest 0.324 4.7 1012 25087
41
+ deu-eng multi30k_task2_test_2016 0.201 3.1 5000 67382
42
+ eng-deu multi30k_task2_test_2016 0.258 2.3 5000 51501
43
+ deu-eng multi30k_test_2016_flickr 0.546 32.2 1000 12955
44
+ eng-deu multi30k_test_2016_flickr 0.582 28.8 1000 12106
45
+ deu-eng multi30k_test_2017_flickr 0.561 32.7 1000 11374
46
+ eng-deu multi30k_test_2017_flickr 0.573 27.6 1000 10755
47
+ deu-eng multi30k_test_2017_mscoco 0.499 25.5 461 5231
48
+ eng-deu multi30k_test_2017_mscoco 0.514 22.0 461 5158
49
+ deu-eng multi30k_test_2018_flickr 0.535 30.0 1071 14689
50
+ eng-deu multi30k_test_2018_flickr 0.547 25.3 1071 13703
51
+ deu-eng newssyscomb2009 0.527 25.4 502 11818
52
+ eng-deu newssyscomb2009 0.504 19.3 502 11271
53
+ deu-eng news-test2008 0.518 23.8 2051 49380
54
+ eng-deu news-test2008 0.492 19.3 2051 47447
55
+ deu-eng newstest2009 0.516 23.4 2525 65399
56
+ eng-deu newstest2009 0.498 18.8 2525 62816
57
+ deu-eng newstest2010 0.546 25.8 2489 61711
58
+ eng-deu newstest2010 0.508 20.7 2489 61503
59
+ deu-eng newstest2011 0.524 23.7 3003 74681
60
+ eng-deu newstest2011 0.493 19.2 3003 72981
61
+ deu-eng newstest2012 0.532 24.8 3003 72812
62
+ eng-deu newstest2012 0.493 19.5 3003 72886
63
+ deu-eng newstest2013 0.548 27.7 3000 64505
64
+ eng-deu newstest2013 0.517 22.5 3000 63737
65
+ deu-eng newstest2014-deen 0.548 27.3 3003 67337
66
+ eng-deu newstest2014-deen 0.532 22.0 3003 62688
67
+ deu-eng newstest2015-deen 0.553 28.6 2169 46443
68
+ eng-deu newstest2015-ende 0.544 25.7 2169 44260
69
+ deu-eng newstest2016-deen 0.596 33.3 2999 64119
70
+ eng-deu newstest2016-ende 0.580 30.0 2999 62669
71
+ deu-eng newstest2017-deen 0.561 29.5 3004 64399
72
+ eng-deu newstest2017-ende 0.535 24.1 3004 61287
73
+ deu-eng newstest2018-deen 0.610 36.1 2998 67012
74
+ eng-deu newstest2018-ende 0.613 35.4 2998 64276
75
+ deu-eng newstest2019-deen 0.582 32.3 2000 39227
76
+ eng-deu newstest2019-ende 0.583 31.2 1997 48746
77
+ deu-eng newstest2020-deen 0.604 32.0 785 38220
78
+ eng-deu newstest2020-ende 0.542 23.9 1418 52383
79
+ deu-eng newstestB2020-deen 0.598 31.2 785 37696
80
+ eng-deu newstestB2020-ende 0.532 23.3 1418 53092
81
+ afr-deu tatoeba-test-v2020-07-28 0.674 48.1 1583 9105
82
+ afr-eng tatoeba-test-v2020-07-28 0.728 58.8 1374 9622
83
+ afr-nld tatoeba-test-v2020-07-28 0.711 54.5 1056 6710
84
+ deu-afr tatoeba-test-v2020-07-28 0.696 52.4 1583 9507
85
+ deu-eng tatoeba-test-v2020-07-28 0.629 44.8 10000 81233
86
+ deu-ltz tatoeba-test-v2020-07-28 0.274 12.7 337 2135
87
+ deu-nds tatoeba-test-v2020-07-28 0.442 18.6 10000 76144
88
+ deu-nld tatoeba-test-v2020-07-28 0.672 48.8 10000 73546
89
+ deu-yid tatoeba-test-v2020-07-28 0.189 2.2 556 3425
90
+ eng-afr tatoeba-test-v2020-07-28 0.735 56.5 1374 10317
91
+ eng-deu tatoeba-test-v2020-07-28 0.591 37.7 10000 83347
92
+ eng-fry tatoeba-test-v2020-07-28 0.376 15.3 205 1529
93
+ eng-ltz tatoeba-test-v2020-07-28 0.280 13.8 283 1733
94
+ eng-nds tatoeba-test-v2020-07-28 0.412 16.6 2500 18264
95
+ eng-nld tatoeba-test-v2020-07-28 0.676 50.1 10000 71436
96
+ eng-yid tatoeba-test-v2020-07-28 0.199 2.2 1168 8094
97
+ fry-eng tatoeba-test-v2020-07-28 0.497 32.1 205 1500
98
+ fry-nld tatoeba-test-v2020-07-28 0.631 42.8 233 1672
99
+ gos-eng tatoeba-test-v2020-07-28 0.315 14.5 1152 5622
100
+ gos-nld tatoeba-test-v2020-07-28 0.405 15.7 1852 9903
101
+ gsw-eng tatoeba-test-v2020-07-28 0.312 14.0 205 990
102
+ hrx-deu tatoeba-test-v2020-07-28 0.484 24.7 471 2805
103
+ hrx-eng tatoeba-test-v2020-07-28 0.362 20.4 221 1235
104
+ ltz-deu tatoeba-test-v2020-07-28 0.553 36.9 337 2144
105
+ ltz-eng tatoeba-test-v2020-07-28 0.479 31.4 283 1751
106
+ ltz-nld tatoeba-test-v2020-07-28 0.518 37.4 273 1567
107
+ nds-deu tatoeba-test-v2020-07-28 0.572 34.5 10000 74571
108
+ nds-eng tatoeba-test-v2020-07-28 0.493 29.9 2500 17589
109
+ nds-nld tatoeba-test-v2020-07-28 0.621 42.3 1657 11490
110
+ nld-afr tatoeba-test-v2020-07-28 0.755 58.8 1056 6823
111
+ nld-deu tatoeba-test-v2020-07-28 0.687 50.4 10000 72438
112
+ nld-eng tatoeba-test-v2020-07-28 0.702 54.5 10000 69848
113
+ nld-fry tatoeba-test-v2020-07-28 0.465 23.7 233 1679
114
+ nld-ltz tatoeba-test-v2020-07-28 0.288 13.3 273 1532
115
+ nld-nds tatoeba-test-v2020-07-28 0.461 21.4 1657 11711
116
+ swg-deu tatoeba-test-v2020-07-28 0.336 8.6 1523 15632
117
+ yid-deu tatoeba-test-v2020-07-28 0.381 17.8 556 3332
118
+ yid-eng tatoeba-test-v2020-07-28 0.337 15.2 1168 7741
119
+ afr-deu tatoeba-test-v2021-03-30 0.674 48.1 1583 9105
120
+ afr-eng tatoeba-test-v2021-03-30 0.728 58.8 1374 9622
121
+ afr-nld tatoeba-test-v2021-03-30 0.711 54.5 1058 6720
122
+ ang-eng tatoeba-test-v2021-03-30 0.215 7.8 200 2132
123
+ deu-afr tatoeba-test-v2021-03-30 0.696 52.4 1583 9507
124
+ deu-eng tatoeba-test-v2021-03-30 0.618 43.4 12664 105121
125
+ deu-ltz tatoeba-test-v2021-03-30 0.278 13.0 350 2227
126
+ deu-nds tatoeba-test-v2021-03-30 0.442 18.6 10000 76144
127
+ deu-nld tatoeba-test-v2021-03-30 0.672 48.8 10124 74568
128
+ deu-yid tatoeba-test-v2021-03-30 0.192 2.2 830 5207
129
+ eng-afr tatoeba-test-v2021-03-30 0.735 56.5 1374 10317
130
+ eng-deu tatoeba-test-v2021-03-30 0.584 36.7 12664 107460
131
+ eng-fry tatoeba-test-v2021-03-30 0.386 16.7 221 1655
132
+ eng-ltz tatoeba-test-v2021-03-30 0.284 14.2 299 1833
133
+ eng-nds tatoeba-test-v2021-03-30 0.412 16.6 2500 18264
134
+ eng-nld tatoeba-test-v2021-03-30 0.668 49.2 11660 83811
135
+ eng-yid tatoeba-test-v2021-03-30 0.206 2.9 1888 12516
136
+ frr-deu tatoeba-test-v2021-03-30 0.199 3.0 279 1886
137
+ fry-eng tatoeba-test-v2021-03-30 0.498 32.2 221 1624
138
+ fry-nld tatoeba-test-v2021-03-30 0.635 43.4 265 1884
139
+ gos-deu tatoeba-test-v2021-03-30 0.369 13.2 210 1175
140
+ gos-eng tatoeba-test-v2021-03-30 0.314 14.4 1193 5819
141
+ gos-nld tatoeba-test-v2021-03-30 0.405 15.7 1852 9903
142
+ gsw-eng tatoeba-test-v2021-03-30 0.311 13.8 210 1021
143
+ hrx-deu tatoeba-test-v2021-03-30 0.484 24.7 473 2813
144
+ hrx-eng tatoeba-test-v2021-03-30 0.364 20.3 223 1249
145
+ ltz-deu tatoeba-test-v2021-03-30 0.556 37.2 350 2229
146
+ ltz-eng tatoeba-test-v2021-03-30 0.487 32.6 299 1845
147
+ ltz-nld tatoeba-test-v2021-03-30 0.531 38.9 306 1763
148
+ nds-deu tatoeba-test-v2021-03-30 0.572 34.5 10000 74571
149
+ nds-eng tatoeba-test-v2021-03-30 0.493 29.9 2500 17589
150
+ nds-nld tatoeba-test-v2021-03-30 0.621 42.3 1657 11490
151
+ nld-afr tatoeba-test-v2021-03-30 0.754 58.7 1058 6833
152
+ nld-deu tatoeba-test-v2021-03-30 0.687 50.5 10124 73449
153
+ nld-eng tatoeba-test-v2021-03-30 0.693 53.5 11660 81885
154
+ nld-fry tatoeba-test-v2021-03-30 0.474 24.4 265 1889
155
+ nld-ltz tatoeba-test-v2021-03-30 0.296 13.5 306 1735
156
+ nld-nds tatoeba-test-v2021-03-30 0.461 21.4 1657 11711
157
+ swg-deu tatoeba-test-v2021-03-30 0.336 8.6 1523 15632
158
+ yid-deu tatoeba-test-v2021-03-30 0.381 17.4 830 5045
159
+ yid-eng tatoeba-test-v2021-03-30 0.333 15.3 1888 11810
160
+ afr-deu tatoeba-test-v2021-08-07 0.674 48.1 1583 9105
161
+ afr-eng tatoeba-test-v2021-08-07 0.728 58.8 1374 9622
162
+ afr-nld tatoeba-test-v2021-08-07 0.711 54.5 1056 6710
163
+ deu-afr tatoeba-test-v2021-08-07 0.696 52.4 1583 9507
164
+ deu-eng tatoeba-test-v2021-08-07 0.609 42.1 17565 149462
165
+ deu-ltz tatoeba-test-v2021-08-07 0.282 13.4 347 2206
166
+ deu-nds tatoeba-test-v2021-08-07 0.442 18.6 9999 76137
167
+ deu-nld tatoeba-test-v2021-08-07 0.672 48.7 10218 75235
168
+ deu-yid tatoeba-test-v2021-08-07 0.198 2.3 853 5355
169
+ eng-afr tatoeba-test-v2021-08-07 0.735 56.5 1374 10317
170
+ eng-deu tatoeba-test-v2021-08-07 0.580 35.9 17565 151568
171
+ eng-fry tatoeba-test-v2021-08-07 0.389 17.2 220 1600
172
+ eng-ltz tatoeba-test-v2021-08-07 0.288 14.0 293 1828
173
+ eng-nds tatoeba-test-v2021-08-07 0.412 16.6 2500 18264
174
+ eng-nld tatoeba-test-v2021-08-07 0.663 48.3 12696 91796
175
+ eng-yid tatoeba-test-v2021-08-07 0.207 2.8 2483 16395
176
+ frr-deu tatoeba-test-v2021-08-07 0.198 3.0 278 1880
177
+ fry-eng tatoeba-test-v2021-08-07 0.500 32.5 220 1573
178
+ fry-nld tatoeba-test-v2021-08-07 0.633 43.1 260 1854
179
+ gos-deu tatoeba-test-v2021-08-07 0.370 12.7 207 1168
180
+ gos-eng tatoeba-test-v2021-08-07 0.316 14.6 1154 5635
181
+ gos-nld tatoeba-test-v2021-08-07 0.405 15.6 1852 9903
182
+ gsw-eng tatoeba-test-v2021-08-07 0.312 14.0 205 990
183
+ hrx-deu tatoeba-test-v2021-08-07 0.484 24.7 471 2805
184
+ hrx-eng tatoeba-test-v2021-08-07 0.362 20.4 221 1235
185
+ ltz-deu tatoeba-test-v2021-08-07 0.556 37.2 347 2208
186
+ ltz-eng tatoeba-test-v2021-08-07 0.485 32.4 293 1840
187
+ ltz-nld tatoeba-test-v2021-08-07 0.534 39.3 292 1685
188
+ nds-deu tatoeba-test-v2021-08-07 0.572 34.5 9999 74564
189
+ nds-eng tatoeba-test-v2021-08-07 0.493 29.9 2500 17589
190
+ nds-nld tatoeba-test-v2021-08-07 0.621 42.3 1657 11490
191
+ nld-afr tatoeba-test-v2021-08-07 0.755 58.8 1056 6823
192
+ nld-deu tatoeba-test-v2021-08-07 0.686 50.4 10218 74131
193
+ nld-eng tatoeba-test-v2021-08-07 0.690 53.1 12696 89978
194
+ nld-fry tatoeba-test-v2021-08-07 0.478 25.1 260 1857
195
+ nld-ltz tatoeba-test-v2021-08-07 0.301 14.1 292 1658
196
+ nld-nds tatoeba-test-v2021-08-07 0.462 21.4 1657 11711
197
+ swg-deu tatoeba-test-v2021-08-07 0.335 8.6 1523 15632
198
+ yid-deu tatoeba-test-v2021-08-07 0.381 17.6 853 5173
199
+ yid-eng tatoeba-test-v2021-08-07 0.324 14.2 2483 15452
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a4586f60a829f1bd8331e4f877b3ddf493c333c821afe8bac9b84847681df2c7
3
  size 161034627
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39d123b7e766718bcf6605ad74a234d47c57867b4e7b1968bede9549faf4e282
3
  size 161034627