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 +1,231 @@
+---
+language:
+- be
+- cs
+- pl
+- ru
+- uk
+- zle
+- zlw
+
+tags:
+- translation
+
+license: cc-by-4.0
+model-index:
+- name: opus-mt-tc-big-zle-zlw
+  results:
+  - task:
+      name: Translation rus-ces
+      type: translation
+      args: rus-ces
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: rus ces devtest
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 23.1
+  - task:
+      name: Translation ukr-ces
+      type: translation
+      args: ukr-ces
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: ukr ces devtest
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 25.1
+  - task:
+      name: Translation bel-pol
+      type: translation
+      args: bel-pol
+    dataset:
+      name: tatoeba-test-v2021-08-07
+      type: tatoeba_mt
+      args: bel-pol
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 47.1
+  - task:
+      name: Translation rus-ces
+      type: translation
+      args: rus-ces
+    dataset:
+      name: tatoeba-test-v2021-08-07
+      type: tatoeba_mt
+      args: rus-ces
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 53.4
+  - task:
+      name: Translation rus-pol
+      type: translation
+      args: rus-pol
+    dataset:
+      name: tatoeba-test-v2021-08-07
+      type: tatoeba_mt
+      args: rus-pol
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 53.7
+  - task:
+      name: Translation ukr-ces
+      type: translation
+      args: ukr-ces
+    dataset:
+      name: tatoeba-test-v2021-08-07
+      type: tatoeba_mt
+      args: ukr-ces
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 58.0
+  - task:
+      name: Translation ukr-pol
+      type: translation
+      args: ukr-pol
+    dataset:
+      name: tatoeba-test-v2021-08-07
+      type: tatoeba_mt
+      args: ukr-pol
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 57.0
+  - task:
+      name: Translation rus-ces
+      type: translation
+      args: rus-ces
+    dataset:
+      name: newstest2013
+      type: wmt-2013-news
+      args: rus-ces
+    metrics:
+       - name: BLEU
+         type: bleu
+         value: 26.0
+---
+# opus-mt-tc-big-zle-zlw
+
+Neural machine translation model for translating from East Slavic languages (zle) to West Slavic languages (zlw).
+
+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: 2022-03-23
+* source language(s): bel rus ukr
+* target language(s): ces pol
+* valid target language labels: >>ces<< >>pol<<
+* model: transformer-big
+* data: opusTCv20210807+bt ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
+* tokenization: SentencePiece (spm32k,spm32k)
+* original model: [opusTCv20210807+bt_transformer-big_2022-03-23.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/zle-zlw/opusTCv20210807+bt_transformer-big_2022-03-23.zip)
+* more information released models: [OPUS-MT zle-zlw README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/zle-zlw/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. `>>ces<<`
+
+## Usage
+
+A short example code:
+
+```python
+from transformers import MarianMTModel, MarianTokenizer
+
+src_text = [
+    ">>pol<< Это метафора.",
+    ">>pol<< Что вы делали?"
+]
+
+model_name = "pytorch-models/opus-mt-tc-big-zle-zlw"
+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:
+#     To metafora.
+#     Co robiliście?
+```
+
+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-zlw")
+print(pipe(">>pol<< Это метафора."))
+
+# expected output: To metafora.
+```
+
+## Benchmarks
+
+* test set translations: [opusTCv20210807+bt_transformer-big_2022-03-23.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/zle-zlw/opusTCv20210807+bt_transformer-big_2022-03-23.test.txt)
+* test set scores: [opusTCv20210807+bt_transformer-big_2022-03-23.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/zle-zlw/opusTCv20210807+bt_transformer-big_2022-03-23.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 |
+|----------|---------|-------|-------|-------|--------|
+| bel-pol | tatoeba-test-v2021-08-07 | 0.65517 | 47.1 | 287 | 1706 |
+| rus-ces | tatoeba-test-v2021-08-07 | 0.69695 | 53.4 | 2934 | 16831 |
+| rus-pol | tatoeba-test-v2021-08-07 | 0.72176 | 53.7 | 3543 | 21505 |
+| ukr-ces | tatoeba-test-v2021-08-07 | 0.73149 | 58.0 | 1787 | 8550 |
+| ukr-pol | tatoeba-test-v2021-08-07 | 0.74649 | 57.0 | 2519 | 13201 |
+| bel-ces | flores101-devtest | 0.41248 | 11.1 | 1012 | 22101 |
+| bel-pol | flores101-devtest | 0.42240 | 10.2 | 1012 | 22520 |
+| rus-ces | flores101-devtest | 0.50971 | 23.1 | 1012 | 22101 |
+| rus-pol | flores101-devtest | 0.48672 | 18.4 | 1012 | 22520 |
+| ukr-ces | flores101-devtest | 0.52482 | 25.1 | 1012 | 22101 |
+| ukr-pol | flores101-devtest | 0.48790 | 18.8 | 1012 | 22520 |
+| rus-ces | newstest2012 | 0.45834 | 18.8 | 3003 | 65456 |
+| rus-ces | newstest2013 | 0.52364 | 26.0 | 3000 | 57250 |
+
+## 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: Thu Mar 24 00:50:29 EET 2022
+* port machine: LM0-400-22516.local

benchmark_results.txt

@@ -0,0 +1,24 @@
+bel-ces	flores101-dev	0.41108	10.4	997	21183
+bel-pol	flores101-dev	0.42162	10.1	997	21684
+rus-ces	flores101-dev	0.51311	23.4	997	21183
+rus-pol	flores101-dev	0.48899	18.4	997	21684
+bel-ces	flores101-devtest	0.41248	11.1	1012	22101
+bel-pol	flores101-devtest	0.42240	10.2	1012	22520
+rus-ces	flores101-devtest	0.50971	23.1	1012	22101
+rus-pol	flores101-devtest	0.48672	18.4	1012	22520
+ukr-ces	flores101-devtest	0.52482	25.1	1012	22101
+ukr-pol	flores101-devtest	0.48790	18.8	1012	22520
+ukr-ces	flores101-dev	0.52364	24.5	997	21183
+ukr-pol	flores101-dev	0.48924	19.0	997	21684
+rus-ces	newstest2012	0.45834	18.8	3003	65456
+rus-ces	newstest2013	0.52364	26.0	3000	57250
+rus-ces	tatoeba-test-v2020-07-28	0.70119	53.8	2500	14266
+ukr-pol	tatoeba-test-v2020-07-28	0.74609	56.9	2500	13098
+bel-pol	tatoeba-test-v2021-03-30	0.65722	47.4	289	1718
+rus-ces	tatoeba-test-v2021-03-30	0.69882	53.6	5060	28961
+ukr-pol	tatoeba-test-v2021-03-30	0.74573	56.9	4977	26119
+bel-pol	tatoeba-test-v2021-08-07	0.65517	47.1	287	1706
+rus-ces	tatoeba-test-v2021-08-07	0.69695	53.4	2934	16831
+rus-pol	tatoeba-test-v2021-08-07	0.72176	53.7	3543	21505
+ukr-ces	tatoeba-test-v2021-08-07	0.73149	58.0	1787	8550
+ukr-pol	tatoeba-test-v2021-08-07	0.74649	57.0	2519	13201

benchmark_translations.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c3fe982f011143d75d5489592f0ecd35b3094198948b001c196cde6eeebd9a4
+size 4290467

config.json

@@ -0,0 +1,45 @@
+{
+  "activation_dropout": 0.0,
+  "activation_function": "relu",
+  "architectures": [
+    "MarianMTModel"
+  ],
+  "attention_dropout": 0.0,
+  "bad_words_ids": [
+    [
+      61592
+    ]
+  ],
+  "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": 61592,
+  "decoder_vocab_size": 61593,
+  "dropout": 0.1,
+  "encoder_attention_heads": 16,
+  "encoder_ffn_dim": 4096,
+  "encoder_layerdrop": 0.0,
+  "encoder_layers": 6,
+  "eos_token_id": 22403,
+  "forced_eos_token_id": 22403,
+  "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": 61592,
+  "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": 61593
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01a1c76789d7cabd1b97d5a838d02b604c79316e4d959bdc645610c65799cb2d
+size 605214531

source.spm

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a940b4375cacfb75b1c7e1d27d4b06a5f130dcd026489e1e693d6fd8ed2259e
+size 1000541

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:f0937cb077654ee99bb4f0897f0b672b1ac19564c84d299b772b95d1f871dd1b
+size 824018

tokenizer_config.json

@@ -0,0 +1 @@
+{"source_lang": "zle", "target_lang": "zlw", "unk_token": "<unk>", "eos_token": "</s>", "pad_token": "<pad>", "model_max_length": 512, "sp_model_kwargs": {}, "separate_vocabs": false, "special_tokens_map_file": null, "name_or_path": "marian-models/opusTCv20210807+bt_transformer-big_2022-03-23/zle-zlw", "tokenizer_class": "MarianTokenizer"}