Initial commit

.gitattributes

@@ -25,3 +25,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zstandard filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.spm filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,243 @@
+---
+language:
+- da
+- gmq
+- nb
+- no
+- ru
+- sv
+- uk
+- zle
+
+tags:
+- translation
+
+license: cc-by-4.0
+model-index:
+- name: opus-mt-tc-big-zle-gmq
+  results:
+  - task:
+      name: Translation rus-dan
+      type: translation
+      args: rus-dan
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: rus dan devtest
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 28.0
+  - task:
+      name: Translation rus-nob
+      type: translation
+      args: rus-nob
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: rus nob devtest
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 20.6
+  - task:
+      name: Translation rus-swe
+      type: translation
+      args: rus-swe
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: rus swe devtest
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 26.4
+  - task:
+      name: Translation ukr-dan
+      type: translation
+      args: ukr-dan
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: ukr dan devtest
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 30.3
+  - task:
+      name: Translation ukr-nob
+      type: translation
+      args: ukr-nob
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: ukr nob devtest
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 21.1
+  - task:
+      name: Translation ukr-swe
+      type: translation
+      args: ukr-swe
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: ukr swe devtest
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 28.8
+  - task:
+      name: Translation rus-dan
+      type: translation
+      args: rus-dan
+    dataset:
+      name: tatoeba-test-v2021-08-07
+      type: tatoeba_mt
+      args: rus-dan
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 59.6
+  - task:
+      name: Translation rus-nob
+      type: translation
+      args: rus-nob
+    dataset:
+      name: tatoeba-test-v2021-08-07
+      type: tatoeba_mt
+      args: rus-nob
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 46.1
+  - task:
+      name: Translation rus-swe
+      type: translation
+      args: rus-swe
+    dataset:
+      name: tatoeba-test-v2021-08-07
+      type: tatoeba_mt
+      args: rus-swe
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 53.3
+---
+# opus-mt-tc-big-zle-gmq
+
+Neural machine translation model for translating from East Slavic languages (zle) to North Germanic languages (gmq).
+
+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).
+
+* 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.)
+
+```
+@inproceedings{tiedemann-thottingal-2020-opus,
+    title = "{OPUS}-{MT} {--} Building open translation services for the World",
+    author = {Tiedemann, J{\"o}rg  and Thottingal, Santhosh},
+    booktitle = "Proceedings of the 22nd Annual Conference of the European Association for Machine Translation",
+    month = nov,
+    year = "2020",
+    address = "Lisboa, Portugal",
+    publisher = "European Association for Machine Translation",
+    url = "https://aclanthology.org/2020.eamt-1.61",
+    pages = "479--480",
+}
+
+@inproceedings{tiedemann-2020-tatoeba,
+    title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
+    author = {Tiedemann, J{\"o}rg},
+    booktitle = "Proceedings of the Fifth Conference on Machine Translation",
+    month = nov,
+    year = "2020",
+    address = "Online",
+    publisher = "Association for Computational Linguistics",
+    url = "https://aclanthology.org/2020.wmt-1.139",
+    pages = "1174--1182",
+}
+```
+
+## Model info
+
+* Release: 2022-03-14
+* source language(s): rus ukr
+* target language(s): dan nob nor swe
+* valid target language labels: >>dan<< >>nob<< >>nor<< >>swe<<
+* model: transformer-big
+* data: opusTCv20210807+pft ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
+* tokenization: SentencePiece (spm32k,spm32k)
+* original model: [opusTCv20210807+pft_transformer-big_2022-03-14.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/zle-gmq/opusTCv20210807+pft_transformer-big_2022-03-14.zip)
+* more information released models: [OPUS-MT zle-gmq README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/zle-gmq/README.md)
+* more information about the model: [MarianMT](https://huggingface.co/docs/transformers/model_doc/marian)
+
+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. `>>dan<<`
+
+## Usage
+
+A short example code:
+
+```python
+from transformers import MarianMTModel, MarianTokenizer
+
+src_text = [
+    ">>dan<< Заўтра ўжо чацвер.",
+    ">>swe<< Том грав з Мері в кішки-мишки."
+]
+
+model_name = "pytorch-models/opus-mt-tc-big-zle-gmq"
+tokenizer = MarianTokenizer.from_pretrained(model_name)
+model = MarianMTModel.from_pretrained(model_name)
+translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))
+
+for t in translated:
+    print( tokenizer.decode(t, skip_special_tokens=True) )
+
+# expected output:
+#     I morgen er det torsdag.
+#     Tom lekte med Mary i katt-möss.
+```
+
+You can also use OPUS-MT models with the transformers pipelines, for example:
+
+```python
+from transformers import pipeline
+pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-zle-gmq")
+print(pipe(">>dan<< Заўтра ўжо чацвер."))
+
+# expected output: I morgen er det torsdag.
+```
+
+## Benchmarks
+
+* test set translations: [opusTCv20210807+pft_transformer-big_2022-03-14.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/zle-gmq/opusTCv20210807+pft_transformer-big_2022-03-14.test.txt)
+* test set scores: [opusTCv20210807+pft_transformer-big_2022-03-14.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/zle-gmq/opusTCv20210807+pft_transformer-big_2022-03-14.eval.txt)
+* benchmark results: [benchmark_results.txt](benchmark_results.txt)
+* benchmark output: [benchmark_translations.zip](benchmark_translations.zip)
+
+| langpair | testset | chr-F | BLEU  | #sent | #words |
+|----------|---------|-------|-------|-------|--------|
+| rus-dan | tatoeba-test-v2021-08-07 | 0.74307 | 59.6 | 1713 | 11746 |
+| rus-nob | tatoeba-test-v2021-08-07 | 0.66376 | 46.1 | 1277 | 11672 |
+| rus-swe | tatoeba-test-v2021-08-07 | 0.69608 | 53.3 | 1282 | 8449 |
+| bel-dan | flores101-devtest | 0.47621 | 13.9 | 1012 | 24638 |
+| bel-nob | flores101-devtest | 0.44966 | 10.8 | 1012 | 23873 |
+| bel-swe | flores101-devtest | 0.47274 | 13.2 | 1012 | 23121 |
+| rus-dan | flores101-devtest | 0.55917 | 28.0 | 1012 | 24638 |
+| rus-nob | flores101-devtest | 0.50724 | 20.6 | 1012 | 23873 |
+| rus-swe | flores101-devtest | 0.55812 | 26.4 | 1012 | 23121 |
+| ukr-dan | flores101-devtest | 0.57829 | 30.3 | 1012 | 24638 |
+| ukr-nob | flores101-devtest | 0.52271 | 21.1 | 1012 | 23873 |
+| ukr-swe | flores101-devtest | 0.57499 | 28.8 | 1012 | 23121 |
+
+## Acknowledgements
+
+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.
+
+## Model conversion info
+
+* transformers version: 4.16.2
+* OPUS-MT git hash: 1bdabf7
+* port time: Wed Mar 23 23:13:54 EET 2022
+* port machine: LM0-400-22516.local

benchmark_results.txt

@@ -0,0 +1,33 @@
+bel-dan	flores101-dev	0.47553	14.2	997	23685
+bel-isl	flores101-dev	0.25300	2.4	997	21857
+bel-nob	flores101-dev	0.44469	11.4	997	23157
+bel-swe	flores101-dev	0.47015	13.5	997	22417
+rus-dan	flores101-dev	0.56354	28.5	997	23685
+rus-isl	flores101-dev	0.26997	4.1	997	21857
+rus-nob	flores101-dev	0.50953	21.2	997	23157
+rus-swe	flores101-dev	0.56381	27.3	997	22417
+bel-dan	flores101-devtest	0.47621	13.9	1012	24638
+bel-isl	flores101-devtest	0.24826	2.6	1012	22834
+bel-nob	flores101-devtest	0.44966	10.8	1012	23873
+bel-swe	flores101-devtest	0.47274	13.2	1012	23121
+rus-dan	flores101-devtest	0.55917	28.0	1012	24638
+rus-isl	flores101-devtest	0.26803	4.1	1012	22834
+rus-nob	flores101-devtest	0.50724	20.6	1012	23873
+rus-swe	flores101-devtest	0.55812	26.4	1012	23121
+ukr-dan	flores101-devtest	0.57829	30.3	1012	24638
+ukr-isl	flores101-devtest	0.27243	4.4	1012	22834
+ukr-nob	flores101-devtest	0.52271	21.1	1012	23873
+ukr-swe	flores101-devtest	0.57499	28.8	1012	23121
+ukr-dan	flores101-dev	0.57956	30.4	997	23685
+ukr-isl	flores101-dev	0.27258	4.3	997	21857
+ukr-nob	flores101-dev	0.52248	21.9	997	23157
+ukr-swe	flores101-dev	0.57381	28.5	997	22417
+rus-dan	tatoeba-test-v2020-07-28	0.74307	59.6	1713	11746
+rus-nob	tatoeba-test-v2020-07-28	0.66376	46.1	1277	11672
+rus-swe	tatoeba-test-v2020-07-28	0.69608	53.3	1282	8449
+rus-dan	tatoeba-test-v2021-03-30	0.74307	59.6	1713	11746
+rus-nob	tatoeba-test-v2021-03-30	0.66376	46.1	1277	11672
+rus-swe	tatoeba-test-v2021-03-30	0.69608	53.3	1282	8449
+rus-dan	tatoeba-test-v2021-08-07	0.74307	59.6	1713	11746
+rus-nob	tatoeba-test-v2021-08-07	0.66376	46.1	1277	11672
+rus-swe	tatoeba-test-v2021-08-07	0.69608	53.3	1282	8449

benchmark_translations.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a489801c3a11cfcec937c2c99d1b8dfac5cc58bd83d740107f1ef4379007b446
+size 5109306

config.json

@@ -0,0 +1,45 @@
+{
+  "activation_dropout": 0.0,
+  "activation_function": "relu",
+  "architectures": [
+    "MarianMTModel"
+  ],
+  "attention_dropout": 0.0,
+  "bad_words_ids": [
+    [
+      61375
+    ]
+  ],
+  "bos_token_id": 0,
+  "classifier_dropout": 0.0,
+  "d_model": 1024,
+  "decoder_attention_heads": 16,
+  "decoder_ffn_dim": 4096,
+  "decoder_layerdrop": 0.0,
+  "decoder_layers": 6,
+  "decoder_start_token_id": 61375,
+  "decoder_vocab_size": 61376,
+  "dropout": 0.1,
+  "encoder_attention_heads": 16,
+  "encoder_ffn_dim": 4096,
+  "encoder_layerdrop": 0.0,
+  "encoder_layers": 6,
+  "eos_token_id": 24306,
+  "forced_eos_token_id": 24306,
+  "init_std": 0.02,
+  "is_encoder_decoder": true,
+  "max_length": 512,
+  "max_position_embeddings": 1024,
+  "model_type": "marian",
+  "normalize_embedding": false,
+  "num_beams": 4,
+  "num_hidden_layers": 6,
+  "pad_token_id": 61375,
+  "scale_embedding": true,
+  "share_encoder_decoder_embeddings": true,
+  "static_position_embeddings": true,
+  "torch_dtype": "float16",
+  "transformers_version": "4.18.0.dev0",
+  "use_cache": true,
+  "vocab_size": 61376
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70e79fad3e862f415be03907ce7b154442e1e2165e8d3bfb6ab05129d4afd502
+size 604325251

source.spm

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd8cc293d1b97fb3347ab64b85da528a533573e10a8a06badb6a5b28b7b20251
+size 1009426

special_tokens_map.json

@@ -0,0 +1 @@
+{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}

target.spm

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:910f577e8872560ea9a4eb5385ef8cda9f9d38adfc3ac6c7fc6d383c03174979
+size 806068

tokenizer_config.json

@@ -0,0 +1 @@
+{"source_lang": "zle", "target_lang": "gmq", "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_transformer-big_2022-03-14/zle-gmq", "tokenizer_class": "MarianTokenizer"}