diff options
author | TharinduDR <rhtdranasinghe@gmail.com> | 2021-02-12 21:19:41 +0300 |
---|---|---|
committer | TharinduDR <rhtdranasinghe@gmail.com> | 2021-02-12 21:19:41 +0300 |
commit | d75da9c76c7981d7709fab78acf97cb1da77a2ee (patch) | |
tree | c14b4b3389883b804fa6c191ff3fbede03d59838 | |
parent | afccb5294f879890c2df6c80e519f7e922a994ea (diff) |
055: Adding word level examples
-rw-r--r-- | docs/architectures/sentence_level_architectures.md (renamed from docs/architectures/architectures.md) | 2 | ||||
-rw-r--r-- | docs/examples/sentence_level_examples.md (renamed from docs/examples/examples.md) | 28 | ||||
-rw-r--r-- | docs/install.md | 6 | ||||
-rw-r--r-- | docs/models/sentence_level_pretrained.md (renamed from docs/models/pretrained.md) | 2 | ||||
-rw-r--r-- | mkdocs.yml | 6 |
5 files changed, 24 insertions, 20 deletions
diff --git a/docs/architectures/architectures.md b/docs/architectures/sentence_level_architectures.md index 042e6e7..d142791 100644 --- a/docs/architectures/architectures.md +++ b/docs/architectures/sentence_level_architectures.md @@ -1,5 +1,5 @@ # TransQuest Architectures -We have introduced two architectures in the TransQuest framework, both relies on the XLM-R transformer model. +We have introduced two architectures for the sentence level QE in the TransQuest framework, both relies on the XLM-R transformer model. ##MonoTransQuest diff --git a/docs/examples/examples.md b/docs/examples/sentence_level_examples.md index dbd47aa..506fd57 100644 --- a/docs/examples/examples.md +++ b/docs/examples/sentence_level_examples.md @@ -1,5 +1,5 @@ # Examples -We have provided several examples on how to use TransQuest in recent WMT sentence-level quality estimation shared tasks. They are included in the repository but are not shipped with the library. Therefore, if you need to run the examples, please clone the repository. +We have provided several examples on how to use TransQuest in recent WMT sentence-level quality estimation shared tasks. They are included in the repository but are not shipped with the library. Therefore, if you need to run the examples, please clone the repository. Please don't use the same environment you used to install transquest to run the examples. Create a new environment. ```bash git clone https://github.com/TharinduDR/TransQuest.git @@ -15,17 +15,17 @@ The participants were predict the direct assessment of a source and a target. Th To run the experiments for each language please run this command from the root directory of TransQuest. ```bash -python -m examples.wmt_2020.<language-pair>.<architecture> +python -m examples.sentence_level.wmt_2020.<language-pair>.<architecture> ``` Language Pair options : ro_en (Romanian-English), ru_en (Russian-English), et_en (Estonian-English), en_zh (English-Chinese), ne_en (Nepalese-English), en_de (English-German), si_en(Sinhala-English) -Architecture Options : trans_quest (MonoTransQuest), siamese_trans_quest (SiameseTransQuest). +Architecture Options : monotransquest (MonoTransQuest), siamesetransquest (SiameseTransQuest). As an example to run the experiments on Romanian-English with MonoTransQuest architecture, run the following command. ```bash -python -m examples.wmt_2020.ro_en.trans_quest +python -m examples.sentence_level.wmt_2020.ro_en.monotransquest ``` ### Results @@ -62,17 +62,17 @@ This task consists predicting Sentence-level HTER (Human Translation Error Rate) To run the experiments for each language please run this command from the root directory of TransQuest. ```bash -python -m examples.wmt_2020_task2.<language-pair>.<architecture> +python -m examples.sentence_level.wmt_2020_task2.<language-pair>.<architecture> ``` Language Pair options : en_zh (English-Chinese), en_de (English-German) -Architecture Options : trans_quest (MonoTransQuest), siamese_trans_quest (SiameseTransQuest). +Architecture Options : monotransquest (MonoTransQuest), siamesetransquest (SiameseTransQuest). As an example to run the experiments on English-Chinese with MonoTransQuest architecture, run the following command. ```bash -python -m examples.wmt_2020_task2.en_zh.trans_quest +python -m examples.sentence_level.wmt_2020_task2.en_zh.monotransquest ``` ### Results @@ -93,17 +93,17 @@ The participating systems are expected to predict the sentence-level HTER score To run the experiments for each language, please run this command from the root directory of TransQuest. ```bash -python -m examples.wmt_2019.<language-pair>.<architecture> +python -m examples.sentence_level.wmt_2019.<language-pair>.<architecture> ``` Language Pair options : en_ru (English-Russian), en_de (English-German) -Architecture Options : trans_quest (MonoTransQuest), siamese_trans_quest (SiameseTransQuest). +Architecture Options : monotransquest (MonoTransQuest), siamesetransquest (SiameseTransQuest). As an example to run the experiments on English-Russian with MonoTransQuest architecture, run the following command. ```bash -python -m examples.wmt_2019.en_ru.trans_quest +python -m examples.sentence_level.wmt_2019.en_ru.trans_quest ``` ### Results @@ -124,23 +124,23 @@ The participating systems are expected to predict the sentence-level HTER score To run the experiments for each language, please run this command from the root directory of TransQuest. If both NMT and SMT is available for a certain language pair, specify that too. ```bash -python -m examples.wmt_2019.<language-pair>.<nmt/smt><architecture> +python -m examples.sentence_level.wmt_2019.<language-pair>.<nmt/smt><architecture> ``` Language Pair options : en_de (English-German) (both NMT and SMT), en_lv(English-Latvian) (both NMT and SMT), en_cs(English-Czech), de_en -Architecture Options : trans_quest (MonoTransQuest), siamese_trans_quest (SiameseTransQuest). +Architecture Options : monotransquest (MonoTransQuest), siamesetransquest (SiameseTransQuest). As an example to run the experiments on English-Latvian NMT with MonoTransQuest architecture, run the following command. ```bash -python -m examples.wmt_2018.en_lv.nmt.trans_quest +python -m examples.sentence_level.wmt_2018.en_lv.nmt.monotransquest ``` To run the English-Czech experiments with MonoTransQuest architecture,, run the following command ```bash -python -m examples.wmt_2018.en_cs.trans_quest +python -m examples.sentence_level.wmt_2018.en_cs.monotransquest ``` diff --git a/docs/install.md b/docs/install.md index d7ca5e5..4bc0cb2 100644 --- a/docs/install.md +++ b/docs/install.md @@ -3,7 +3,7 @@ You first need to install PyTorch. The recommended PyTorch version is 1.6. Please refer to [PyTorch installation page](https://pytorch.org/get-started/locally/#start-locally) regarding the specific install command for your platform. When PyTorch has been installed, you can install TransQuest from source or from pip. - +If you are training models, we highly recommend using a GPU. We used a NVIDIA TESLA K80 GPU to train the models. ### From pip @@ -18,3 +18,7 @@ git clone https://github.com/TharinduDR/TransQuest.git cd TransQuest pip install -r requirements.txt ``` + + + + diff --git a/docs/models/pretrained.md b/docs/models/sentence_level_pretrained.md index 372cd1a..95211b8 100644 --- a/docs/models/pretrained.md +++ b/docs/models/sentence_level_pretrained.md @@ -1,4 +1,4 @@ -# Pre-trained Models +# Sentence Level Pre-trained Models We have released several pre-trained TransQuest models on two aspects in sentence-level quality estimation. We will be keep releasing new models. So please keep in touch. ## Predicting HTER @@ -2,9 +2,9 @@ site_name: TransQuest nav: - Home: index.md - - TransQuest Architectures: architectures.md - - Pre-trained Models: pretrained.md - - Examples: examples.md + - TransQuest Architectures: sentence_level_architectures.md + - Pre-trained Models: sentence_level_pretrained.md + - Examples: sentence_level_examples.md - Contact: contact.md theme: material |