tiedeman commited on
Commit
746b42a
1 Parent(s): 4fabfef

Initial commit

Browse files
.gitattributes CHANGED
@@ -25,3 +25,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
25
  *.zip filter=lfs diff=lfs merge=lfs -text
26
  *.zstandard filter=lfs diff=lfs merge=lfs -text
27
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
25
  *.zip filter=lfs diff=lfs merge=lfs -text
26
  *.zstandard filter=lfs diff=lfs merge=lfs -text
27
  *tfevents* filter=lfs diff=lfs merge=lfs -text
28
+ *.spm filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,114 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - fi
4
+ - uk
5
+
6
+ tags:
7
+ - translation
8
+
9
+ license: cc-by-4.0
10
+ model-index:
11
+ - name: opus-mt-tc-base-uk-fi
12
+ results:
13
+ ---
14
+ # opus-mt-tc-base-uk-fi
15
+
16
+ Neural machine translation model for translating from Ukrainian (uk) to Finnish (fi).
17
+
18
+ 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).
19
+
20
+ * 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.)
21
+
22
+ ```
23
+ @inproceedings{tiedemann-thottingal-2020-opus,
24
+ title = "{OPUS}-{MT} {--} Building open translation services for the World",
25
+ author = {Tiedemann, J{\"o}rg and Thottingal, Santhosh},
26
+ booktitle = "Proceedings of the 22nd Annual Conference of the European Association for Machine Translation",
27
+ month = nov,
28
+ year = "2020",
29
+ address = "Lisboa, Portugal",
30
+ publisher = "European Association for Machine Translation",
31
+ url = "https://aclanthology.org/2020.eamt-1.61",
32
+ pages = "479--480",
33
+ }
34
+
35
+ @inproceedings{tiedemann-2020-tatoeba,
36
+ title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
37
+ author = {Tiedemann, J{\"o}rg},
38
+ booktitle = "Proceedings of the Fifth Conference on Machine Translation",
39
+ month = nov,
40
+ year = "2020",
41
+ address = "Online",
42
+ publisher = "Association for Computational Linguistics",
43
+ url = "https://aclanthology.org/2020.wmt-1.139",
44
+ pages = "1174--1182",
45
+ }
46
+ ```
47
+
48
+ ## Model info
49
+
50
+ * Release: 2022-03-17
51
+ * source language(s): ukr
52
+ * target language(s): fin
53
+ * model: transformer-align (base)
54
+ * data: opusTCv20210807+pft+pbt ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
55
+ * tokenization: SentencePiece (spm32k,spm32k)
56
+ * original model: [opusTCv20210807+pft+pbt_transformer-align_2022-03-17.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/ukr-fin/opusTCv20210807+pft+pbt_transformer-align_2022-03-17.zip)
57
+ * more information released models: [OPUS-MT ukr-fin README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/ukr-fin/README.md)
58
+
59
+ ## Usage
60
+
61
+ A short example code:
62
+
63
+ ```python
64
+ from transformers import MarianMTModel, MarianTokenizer
65
+
66
+ src_text = [
67
+ "Африка є колискою людства.",
68
+ "Один, два, три, чотири, п'ять, шість, сім, вісім, дев'ять, десять."
69
+ ]
70
+
71
+ model_name = "pytorch-models/opus-mt-tc-base-uk-fi"
72
+ tokenizer = MarianTokenizer.from_pretrained(model_name)
73
+ model = MarianMTModel.from_pretrained(model_name)
74
+ translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))
75
+
76
+ for t in translated:
77
+ print( tokenizer.decode(t, skip_special_tokens=True) )
78
+
79
+ # expected output:
80
+ # Afrikka on ihmiskunnan kehto.
81
+ # Yksi, kaksi, kolme, neljä, viisi, kuusi, seitsemän, kahdeksan, yhdeksän, kymmenen.
82
+ ```
83
+
84
+ You can also use OPUS-MT models with the transformers pipelines, for example:
85
+
86
+ ```python
87
+ from transformers import pipeline
88
+ pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-base-uk-fi")
89
+ print(pipe("Африка є колискою людства."))
90
+
91
+ # expected output: Afrikka on ihmiskunnan kehto.
92
+ ```
93
+
94
+ ## Benchmarks
95
+
96
+ * test set translations: [opusTCv20210807+pft+pbt_transformer-align_2022-03-17.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/ukr-fin/opusTCv20210807+pft+pbt_transformer-align_2022-03-17.test.txt)
97
+ * test set scores: [opusTCv20210807+pft+pbt_transformer-align_2022-03-17.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/ukr-fin/opusTCv20210807+pft+pbt_transformer-align_2022-03-17.eval.txt)
98
+ * benchmark results: [benchmark_results.txt](benchmark_results.txt)
99
+ * benchmark output: [benchmark_translations.zip](benchmark_translations.zip)
100
+
101
+ | langpair | testset | chr-F | BLEU | #sent | #words |
102
+ |----------|---------|-------|-------|-------|--------|
103
+ | ukr-fin | flores101-devtest | 0.54827 | 19.6 | 1012 | 18781 |
104
+
105
+ ## Acknowledgements
106
+
107
+ 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.
108
+
109
+ ## Model conversion info
110
+
111
+ * transformers version: 4.16.2
112
+ * OPUS-MT git hash: f084bad
113
+ * port time: Wed Mar 23 21:02:02 EET 2022
114
+ * port machine: LM0-400-22516.local
benchmark_results.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ ukr-fin flores101-devtest 0.54827 19.6 1012 18781
2
+ ukr-fin flores101-dev 0.54009 19.6 997 17938
benchmark_translations.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16fae1a5d2e11fb4740118912061e49d9e0179d10a106830853f81a658f145a3
3
+ size 370567
config.json ADDED
@@ -0,0 +1,45 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_dropout": 0.0,
3
+ "activation_function": "swish",
4
+ "architectures": [
5
+ "MarianMTModel"
6
+ ],
7
+ "attention_dropout": 0.0,
8
+ "bad_words_ids": [
9
+ [
10
+ 60734
11
+ ]
12
+ ],
13
+ "bos_token_id": 0,
14
+ "classifier_dropout": 0.0,
15
+ "d_model": 512,
16
+ "decoder_attention_heads": 8,
17
+ "decoder_ffn_dim": 2048,
18
+ "decoder_layerdrop": 0.0,
19
+ "decoder_layers": 6,
20
+ "decoder_start_token_id": 60734,
21
+ "decoder_vocab_size": 60735,
22
+ "dropout": 0.1,
23
+ "encoder_attention_heads": 8,
24
+ "encoder_ffn_dim": 2048,
25
+ "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 6,
27
+ "eos_token_id": 23096,
28
+ "forced_eos_token_id": 23096,
29
+ "init_std": 0.02,
30
+ "is_encoder_decoder": true,
31
+ "max_length": 512,
32
+ "max_position_embeddings": 512,
33
+ "model_type": "marian",
34
+ "normalize_embedding": false,
35
+ "num_beams": 4,
36
+ "num_hidden_layers": 6,
37
+ "pad_token_id": 60734,
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": 60735
45
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8055d33b85dd76f306add7fbe6c70e3c184c5f7f7bc285cb240d4c84e5506026
3
+ size 212875395
source.spm ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93f1b326d44225aa6fca7cb8cbfd1f9b030c6cfc13ea0d832b4c15f09f788ca5
3
+ size 1000715
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 sha256:94886ee0cf3259fe334f705401ea10e9729cad2358dca396f6d4cf06f82940c1
3
+ size 834304
tokenizer_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"source_lang": "uk", "target_lang": "fi", "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+pft+pbt_transformer-align_2022-03-17/uk-fi", "tokenizer_class": "MarianTokenizer"}
vocab.json ADDED
The diff for this file is too large to render. See raw diff