README.md

---
license: apache-2.0
language:
- bg
- ca
- cs
- cy
- da
- de
- el
- en
- es
- et
- eu
- fi
- fr
- ga
- gl
- hr
- hu
- it
- lt
- lv
- mt
- nl
- nb
- 'no'
- nn
- oc
- pl
- pt
- ro
- ru
- sl
- sk
- sr
- sv
- uk
- ast
- an
---


![](./images/salamandra_header.png)

# Salamandra Model Card

## How to use

> [!IMPORTANT]
> This version of Salamandra is tailored exclusively for translation tasks. It lacks chat capabilities and has not been trained with any chat instructions.


The instruction-following models use the commonly adopted ChatML template:

```
<|im_start|>system
{SYSTEM PROMPT}<|im_end|>
<|im_start|>user
{USER PROMPT}<|im_end|>
<|im_start|>assistant
{MODEL RESPONSE}<|im_end|>
<|im_start|>user
[...]
```

The easiest way to apply it is by using the tokenizer's built-in functions, as shown in the following snippet.

```python
from datetime import datetime
from transformers import AutoTokenizer, AutoModelForCausalLM
import transformers
import torch

model_id = "/gpfs/projects/bsc88/mt_translation/instructed_models/salamandraTA7b_instruct_mixture1/checkpoint-510"

source = 'Spanish'
target = 'Catalan'
sentence = "Pensando en ti y en este amor que parte mi universo en dos y que llega del olvido hasta mi propia voz y araña mi pasado sin pedir perdón"
 
text = f"Translate the following text from {source} into {target}.\n{source}: {sentence} \n{target}:"


tokenizer = AutoTokenizer.from_pretrained(model_id)

stop_sequence = '<|im_end|>'
eos_tokens = [tokenizer.eos_token_id,tokenizer.convert_tokens_to_ids(stop_sequence)]

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype=torch.bfloat16
  )

message = [ { "role": "user", "content": text } ]
date_string = datetime.today().strftime('%Y-%m-%d')

prompt = tokenizer.apply_chat_template(
    message,
    tokenize=False,
    add_generation_prompt=True,
    date_string=date_string
)

inputs = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt")
input_length = inputs.shape[1]
outputs = model.generate(input_ids=inputs.to(model.device), 
                         max_new_tokens=400,
                         early_stopping=True,
                         eos_token_id=eos_tokens, 
                         pad_token_id=tokenizer.eos_token_id,
                         num_beams=5)

print(tokenizer.decode(outputs[0, input_length:], skip_special_tokens=True))
# Pensant en tu i en aquest amor que parteix el meu univers en dos i que arriba des de l'oblit fins a la meva pròpia veu i esgarrapa el meu passat sense demanar perdó
```


Using this template, each turn is preceded by a `<|im_start|>` delimiter and the role of the entity 
(either `user`, for content supplied by the user, or `assistant` for LLM responses), and finished with the `<|im_end|>` token.


## Data

### Pretraining Data

The training corpus consists of 70+XX billion tokens of Catalan-, Spanish-centric, and English-centric parallel data, including all of the official European languages plus Catalan, Basque, 
Galician, Asturian, Aragonese and Aranese. It amounts to 6,574,251,526 parallel sentence pairs. 

This highly multilingual corpus is predominantly composed of data sourced from [OPUS](https://opus.nlpl.eu/), with additional data taken from the [NTEU project](https://nteu.eu/), Project Aina’s existing corpora, and our own unpublished datasets. 
Where little parallel Catalan <-> xx data could be found, synthetic Catalan data was generated from the Spanish side of the collected Spanish <-> xx corpora using 
[Projecte Aina’s Spanish-Catalan model](https://huggingface.co/projecte-aina/aina-translator-es-ca). The final distribution of languages was as below:

![](./treemap.png)

Click the expand button below to see the full list of corpora included in the training data.

<details>
<summary>Data Sources</summary>
 
| Dataset                                   	| Ca-xx Languages                                                                                                	|  Es-xx Langugages                             | En-xx Languages |
|-----------------------------------------------|----------------------------------------------------------------|-----------------------------------------------|----------------------------------------------------------------|
|[AINA](https://huggingface.co/projecte-aina) | en     |       |       |
|ARANESE-SYNTH-CORPUS-BSC                         | arn   |      |         |
|BOUA-BSC             |     | val |        |
|[BOUMH](https://github.com/transducens/PILAR/tree/main/valencian/BOUMH) |          | val   |           |
|[BOUA-PILAR](https://github.com/transducens/PILAR/tree/main/valencian/BOUA)  |          | val |       |
|[CCMatrix](https://opus.nlpl.eu/CCMatrix/corpus/version/CCMatrix)		|eu			|		| ga | 
|[DGT](https://opus.nlpl.eu/DGT/corpus/version/DGT)			|			|bg,cs,da,de,el	,et,fi,fr,ga,hr,hu,lt,lv,mt,nl,pl,pt,ro,sk,sl,sv	|    da,et,ga,hr,hu,lt,lv,mt,sh,sl|
|DOGV-BSC                 |      |  val    |         |
|[DOGV-PILAR](https://github.com/transducens/PILAR/tree/main/valencian/DOGV-html) |           | val |            | 
|[ELRC-EMEA](https://opus.nlpl.eu/ELRC-EMEA/corpus/version/ELRC-EMEA)		|			|bg,cs,da,hu,lt,lv,mt,pl,ro,sk,sl		| et,hr,lv,ro,sk,sl |
|[EMEA](https://opus.nlpl.eu/EMEA/corpus/version/EMEA)			|			|bg,cs,da,el,fi,hu,lt,mt,nl,pl,ro,sk,sl,sv		|    et,mt  |
|[EUBookshop](https://opus.nlpl.eu/EUbookshop/corpus/version/EUbookshop)		|lt,pl,pt			|cs,da,de,el,fi,fr,ga,it,lv,mt,nl,pl,pt,ro,sk,sl,sv		|cy,ga|
|[Europarl](https://opus.nlpl.eu/Europarl/corpus/version/Europarl)		|			|bg,cs,da,el,en,fi,fr,hu,lt,lv,nl,pl,pt	,ro,sk,sl,sv	| |
|[Europat](https://opus.nlpl.eu/EuroPat/corpus/version/EuroPat)		|			|en,hr		| no  |
|[GAITU](https://gaitu.eus/) | | | eu|
|[KDE4](https://opus.nlpl.eu/KDE4/corpus/version/KDE4)			|bg,cs,da,de,el	,et,eu,fi,fr,ga,gl,hr,it,lt,lv,nl,pl,pt,ro,sk,sl,sv	|bg,ga,hr	|cy,ga,nn,oc |
|[GlobalVoices](https://opus.nlpl.eu/GlobalVoices/corpus/version/GlobalVoices)		| bg,de,fr,it,nl,pl,pt	|bg,de,fr,pt		|  |
|[GNOME](https://opus.nlpl.eu/GNOME/corpus/version/GNOME)		|eu,fr,ga,gl,pt		|ga		|cy,ga,nn|  
|[JRC-Arquis](https://opus.nlpl.eu/JRC-Acquis/corpus/version/JRC-Acquis)		|			|cs,da,et,fr,lt,lv,mt,nl,pl	,ro,sv|	 et  |
|LES-CORTS-VALENCIANES-BSC  |            | val            |           |
|[MaCoCu](https://opus.nlpl.eu/MaCoCu/corpus/version/MaCoCu)                    | en     |     | hr,mt,uk   |
|[MultiCCAligned](https://opus.nlpl.eu/JRC-Acquis/corpus/version/JRC-Acquis)	|bg,cs,de,el,et,fi,fr,hr,hu,it,lt,lv,nl,pl,ro,sk,sv	|bg,fi,fr,hr,it,lv,nl,pt		|bg,cy,da,et,fi,hr,hu,lt,lv,no,sl,sr,uk|
|[MultiHPLT](https://opus.nlpl.eu/MultiHPLT/corpus/version/MultiHPLT)		|en, et,fi,ga,hr,mt		|		|fi,ga,gl,hr,mt,nn,sr |
|[MultiParaCrawl](https://opus.nlpl.eu/MultiParaCrawl/corpus/version/MultiParaCrawl)	|bg,da		|de,en,fr,ga,hr,hu,it,mt,pt		|bg,cs,da,de,el,et,fi,fr,ga,hr,hu,lt,lv,mt,nn,pl,ro,sk,sl,uk|
|[MultiUN](https://opus.nlpl.eu/MultiUN/corpus/version/MultiUN)		|			|fr	|	|	
|[News-Commentary](https://opus.nlpl.eu/News-Commentary/corpus/version/News-Commentary) 	|		|fr		|  |
|[NLLB](https://opus.nlpl.eu/NLLB/corpus/version/NLLB)			|bg,da,el,en,et,fi,fr,gl,hu,it	,lt,lv,pt,ro,sk,sl	|bg,cs,da,de,el	,et,fi,fr,hu,it,lt,lv,nl,pl,pt	,ro,sk,sl,sv| bg,cs,cy,da,de,el,et,fi,fr,ga,hr,hu,it,lt,lv,mt,nl,no,oc,pl,pt,ro,ru,sk,sl,sr,sv,uk|
|[NÓS](https://zenodo.org/records/7675110)                 |               |               |    gl      |
|[NÓS-SYN](https://zenodo.org/records/7685180)            |                |               |   gl       |
|[NTEU](https://www.elrc-share.eu/repository/search/?q=NTEU)			|			|bg,cs,da,de,el,en,et,fi,fr,ga,hr,hu,it,lt,lv,mt,nl,pl,pt,ro,sk,sl,sv	|        da,et,ga,hr,lt,lv,mt,ro,sk,sl,sv     |	 
|[OpenSubtitles](https://opus.nlpl.eu/OpenSubtitles/corpus/version/OpenSubtitles) 	|bg,cs,da,de,el	,et,eu,fi,gl,hr,hu,lt,lv,nl,pl,pt,ro,sk,sl,sv	|da,de,fi,fr,hr,hu,it,lv,nl		| bg,cs,de,el,et,hr,fi,fr,hr,hu,no,sl,sr|
|[OPUS-100](https://opus.nlpl.eu/opus-100.php)   | en |  | gl |
|[StanfordNLP-NMT](https://opus.nlpl.eu/StanfordNLP-NMT/corpus/version/StanfordNLP-NMT) | | |cs |
|[Tatoeba](https://opus.nlpl.eu/Tatoeba/corpus/version/Tatoeba)		|de,pt			|pt		|   |
|[TildeModel](https://opus.nlpl.eu/TildeMODEL/corpus/version/TildeMODEL)		|			|bg		| et,hr,lt,lv,mt |
|[UNPC](https://opus.nlpl.eu/UNPC/corpus/version/UNPC)			|			|en,fr		| ru  |	
|[VALENCIAN-AUTH](https://github.com/transducens/PILAR/tree/main/valencian/Generalitat)  |          |    val     |          |
|[VALENCIAN-SYNTH](https://github.com/transducens/PILAR/tree/main/valencian/Generalitat)  |      | val |     |
|[WikiMatrix](https://opus.nlpl.eu/WikiMatrix/corpus/version/WikiMatrix)		|bg,cs,da,de,el	,et,eu,fi,fr,gl,hr,hu,it,lt,nl,pl,pt,ro,sk,sl,sv	|bg,en,fr,hr,it,pt		| oc,sh |
|[Wikimedia](https://opus.nlpl.eu/wikimedia/corpus/version/wikimedia) | | |cy,nn |
|[XLENT](https://opus.nlpl.eu/XLEnt/corpus/version/XLEnt)		|eu,ga,gl			|ga		|cy,et,ga,gl,hr,oc,sh|




</details>


We provide an extense Datasheet section following the best practices defined by [(Gebru et al., 2021)](https://arxiv.org/pdf/1803.09010).

<details>
<summary>Datasheet</summary>

#### Motivation

**For what purpose was the dataset created? Was there a specific task in mind? Was there a specific gap that needed to be filled? Please provide a description.**

The purpose of creating this dataset is to pre-train multilingual models on parallel data in a large number of European languages, with Spanish and Catalan as the pivot languages. We have found that there is a lack of high quality parallel data in the scale necessary for training models, particularly between mid to low resource languages, and so in this dataset we have attempted to compile all publicly available resources for the included smaller languages, in addition to creating additional resources for Catalan as the pivot language.

**Who created the dataset (e.g., which team, research group) and on behalf of which entity (e.g., company, institution, organization)?**

The dataset has been created by the Machine Translation sub-group of the Language Technologies unit (LangTech) of the Barcelona Supercomputing Center - Centro Nacional de
Supercomputación (BSC-CNS), which aims to advance the field of natural language processing through cutting-edge research and development
and the use of HPC. In particular, the main contributors were Audrey Mash and Francesca De Luca Fornaciari.

**Who funded the creation of the dataset? If there is an associated grant, please provide the name of the grantor and the grant name and number.**

This work/research has been promoted and financed by the Government of Catalonia through the [Aina project](https://projecteaina.cat/).

#### Composition

**What do the instances that comprise the dataset represent (e.g., documents, photos, people, countries)? Are there multiple types of instances (e.g., movies, users, and ratings; people and interactions between them; nodes and edges)? Please provide a description.**

The dataset consists entirely of parallel text separated at sentence level. Specifically, data was mainly sourced from the following databases and
repositories:
- **[Opus](https://opus.nlpl.eu/):** Repository which aims to provide freely available parallel datasets in order to advance work in computational linguistics and automatic translation. 
- **[ELRC-SHARE](https://www.elrc-share.eu/):** Repository used for documenting, storing, browsing and accessing Language Resources that are collected through the European Language Resource Coordination.

**How many instances are there in total (of each type, if appropriate)?**

The dataset contains a diverse range of sentence pairs across multiple languages. 36.02% of the data is parallel with Catalan, 27.59% is parallel with Spanish and 0.37% is parallel with English.

**Does the dataset contain all possible instances or is it a sample (not necessarily random) of instances from a larger set? If the dataset is a sample, then what is the larger set? Is the sample representative of the larger set (e.g., geographic coverage)? If so, please describe how this representativeness was validated/verified. If it is not representative of the larger set, please describe why not (e.g., to cover a more diverse range of instances, because instances were withheld or unavailable).**

The dataset is a sample from various sources. Language pairs which had fewer than 100 million parallel sentence pairs after filtering and cleaning were taken 
in their entirety. A sample of 100 million sentence pairs was taken from language pairs which had more data than this after preprocessing. All sampling was random. 
Where very little data existed between Catalan and the target language, synthetic Catalan data was created in order to increase the sample size. 
This was done using [Projecte Aina’s Spanish-Catalan model](https://huggingface.co/projecte-aina/aina-translator-es-ca).

**What data does each instance consist of? “Raw” data (e.g., unprocessed text or images) or features? In either case, please provide a description.**

Each instance consists of a parallel sentence pair processed for deduplication, language identification, and language alignment. 

**Is there a label or target associated with each instance? If so, please provide a description.**

Each instance is labelled with the two languages present in the sentence pair.

**Is any information missing from individual instances? If so, please provide a description, explaining why this information is missing (e.g., because it was unavailable). This does not include intentionally removed information, but might include, e.g., redacted text.**

No significant information is missing from the instances.

**Are relationships between individual instances made explicit (e.g., users’ movie ratings, social network links)? If so, please describe how these relationships are made explicit.**

Instances are related through shared language identifiers.

**Are there recommended data splits (e.g., training, development/validation, testing)? If so, please provide a description of these splits, explaining the rationale behind them.**

The dataset is split randomly into training, validation, and test sets.

**Are there any errors, sources of noise, or redundancies in the dataset? If so, please provide a description.**

Despite filtering for alignment and language identification, a small number of misaligned sentence pairs and incorrectly labelled languages may remain present in the data. The thresholds chosen for this task aim to achieve an optimal balance, prioritising higher accuracy.

**Is the dataset self-contained, or does it link to or otherwise rely on external resources (e.g., websites, tweets, other datasets)? If it links to or relies on external resources, a) are there guarantees that they will exist, and remain constant, over time; b) are there official archival versions of the complete dataset (i.e., including the external resources as they existed at the time the dataset was created); c) are there any restrictions (e.g., licenses, fees) associated with any of the external resources that might apply to a dataset consumer? Please provide descriptions of all external resources and any restrictions associated with them, as well as links or other access points, as appropriate.**

The dataset is self-contained and does not rely on external resources.

**Does the dataset contain data that might be considered confidential (e.g., data that is protected by legal privilege or by doctor–patient confidentiality, data that includes the content of individuals’ non-public communications)? If so, please provide a description.**

The dataset does not contain confidential data.

**Does the dataset contain data that, if viewed directly, might be offensive, insulting, threatening, or might otherwise cause anxiety? If so, please describe why. If the dataset does not relate to people, you may skip the remaining questions in this section.**

The dataset includes web-crawled content, which may overrepresent pornographic material across languages (Kreutzer et al., 2022). We have performed no filtering for toxic material.

**Does the dataset identify any subpopulations (e.g., by age, gender)? If so, please describe how these subpopulations are identified and provide a description of their respective distributions within the dataset.**

The dataset does not explicitly identify any subpopulations.

**Is it possible to identify individuals (i.e., one or more natural persons), either directly or indirectly (i.e., in combination with other data) from the dataset? If so, please describe how.**

Web-sourced instances in the dataset may contain personally identifiable information (PII) that is publicly available on the Web, such as
names, IP addresses, email addresses, and phone numbers. While it would be possible to indirectly identify individuals through the
combination of multiple data points, the nature and scale of web data makes it difficult to parse such information. 

**Does the dataset contain data that might be considered sensitive in any way? If so, please provide a description.**

Given that the dataset includes web-sourced content and other publicly available documents, instances may inadvertently reveal financial
information, health-related details, or forms of government identification, such as social security numbers (Subramani et al., 2023),
especially if the content originates from less-regulated sources or user-generated platforms.

#### Collection Process

**How was the data collected?**

This dataset is constituted by combining several sources, all of which take the form of web-sourced datasets with some preprocessing available under permissive license (p.e. Common Crawl).

**What mechanisms or procedures were used to collect the data? How were these mechanisms or procedures validated?**

All datasets were acquired through open direct download and validated with data integrity tests. 

**If the dataset is a sample from a larger set, what was the sampling strategy?**

The sampling strategy was to use the whole dataset resulting from the filtering explained in the ‘preprocessing/cleaning/labelling’ section,
with the particularity that language pairs consisting of over 100 million sentence pairs were randomly sampled down to 100 million.

**Who was involved in the data collection process and how were they compensated?**

This data is generally extracted, filtered and sampled by automated processes. The code required to run these processes has been developed
entirely by members of the LangTech data team, or otherwise obtained from open-source software. Furthermore, there has been no monetary
consideration for acquiring data from suppliers.

**Over what timeframe was the data collected? Does this timeframe match the creation timeframe of the data associated with the instances? If not, please describe the timeframe in which the data associated with the instances was created.**

Data were acquired and processed from April 2023 to August 2024. However, as mentioned, much data has been obtained from open projects such
as Common Crawl, which contains data from 2014, so it is the end date (04/2024) rather than the start date that is important.

**Were any ethical review processes conducted? If so, please provide a description of these review processes, including the outcomes, as well as a link or other access point to any supporting documentation.**

No particular ethical review process has been carried out as the data is mostly open and not particularly sensitive. However, we have an
internal evaluation team and a bias team to monitor ethical issues. In addition, we work closely with ‘Observatori d'Ètica en Intel·ligència
Artificial’ (OEIAC) and  ‘Agencia Española de Supervisión de la Inteligencia Artificial’ (AESIA) to audit the processes we carry out from an
ethical and legal point of view, respectively.

#### Preprocessing

**Was any preprocessing/cleaning/labeling of the data done? If so, please provide a description. If not, you may skip the remaining questions in this section.**

All data was filtered according to two specific criteria:
- Alignment - sentence level alignments were calculated using [LaBSE](https://huggingface.co/sentence-transformers/LaBSE) and sentence pairs with a score below 0.75 were discarded.
- Language identification - The probability of being the target language was calculated using either [Idiomata Cognitor](https://github.com/transducens/idiomata_cognitor) or [Lingua.py](https://github.com/pemistahl/lingua-py) and sentences identified as unlikely to be the correct language were filtered out. Thresholds varied by language. 

**Was the “raw” data saved in addition to the preprocessed/cleaned/labeled data? If so, please provide a link or other access point to the “raw” data.**

The original raw data was kept on the BSC servers but is not publicly available.

**Is the software that was used to preprocess/clean/label the data available? If so, please provide a link or other access point.**

No, our internal cleaning pipeline for parallel data has not been made publicly available.

#### Uses

**Has the dataset been used for any tasks already? If so, please provide a description.**

Pre-train the SalamandraTA model family.

**What (other) tasks could the dataset be used for?**

The data can be used primarily to pre-train other Machine Translation models.

**Is there anything about the composition of the dataset or the way it was collected and preprocessed/cleaned/labeled that might impact future uses? Is there anything a dataset consumer could do to mitigate these risks or harms?**
 
Web-crawled content is over-represented with standard language varieties, impacting language model performance for minority languages.
Language diversity in data is crucial to avoid bias, especially in encoding non-standard dialects, preventing the exclusion of demographic
groups. Moreover, despite legal uncertainties in web-scraped data, we prioritize permissive licenses and privacy protection measures,
acknowledging the challenges posed by personally identifiable information (PII) within large-scale datasets. Our ongoing efforts aim to
address privacy concerns and contribute to a more inclusive linguistic dataset.

**Are there tasks for which the dataset should not be used?**

-

#### Distribution

**Will the dataset be distributed to third parties outside of the entity on behalf of which the dataset was created? If so, please provide a description.**

The dataset will not be released or distributed to third parties. Any related question to distribution is omitted in this section.

#### Maintenance

**Who will be supporting/hosting/maintaining the dataset?**

The dataset will be hosted by the Language Technologies unit (LangTech) of the Barcelona Supercomputing Center (BSC). The team will ensure
regular updates and monitor the dataset for any issues related to content integrity, legal compliance, and bias for the sources they are
responsible for.

**How can the owner/curator/manager of the dataset be contacted?**

The data owner may be contacted with the email address langtech@bsc.es.

**Will the dataset be updated?**

The dataset will not be updated.

**If the dataset relates to people, are there applicable limits on the retention of the data associated with the instances? If so, please describe these limits and explain how they will be enforced.**

The dataset does not keep sensitive data that could allow direct identification of individuals, apart from the data that is publicly
available in web-sourced content. Due to the sheer volume and diversity of web data, it is not feasible to notify individuals or manage data
retention on an individual basis. However, efforts are made to mitigate the risks associated with sensitive information through
pre-processing and filtering to remove identifiable or harmful content. Despite these measures, vigilance is maintained to address potential
privacy and ethical issues.

**Will older versions of the dataset continue to be supported/hosted/maintained? If so, please describe how. If not, please describe how its obsolescence will be communicated to dataset consumers.**

Since the dataset will not be updated, only the final version will be kept.

**If others want to extend/augment/build on/contribute to the dataset, is there a mechanism for them to do so?**

The dataset does not allow for external contributions.

</details>

<details>
<summary>References</summary>

- Aulamo, M., Sulubacak, U., Virpioja, S., & Tiedemann, J. (2020). OpusTools and Parallel Corpus Diagnostics. In N. Calzolari, F. Béchet, P. Blache, K. Choukri, C. Cieri, T. Declerck, S. Goggi, H. Isahara, B. Maegaard, J. Mariani, H. Mazo, A. Moreno, J. Odijk, & S. Piperidis (Eds.), Proceedings of the Twelfth Language Resources and Evaluation Conference (pp. 3782–3789). European Language Resources Association. https://aclanthology.org/2020.lrec-1.467
- Chaudhary, V., Tang, Y., Guzmán, F., Schwenk, H., & Koehn, P. (2019). Low-Resource Corpus Filtering Using Multilingual Sentence Embeddings. In O. Bojar, R. Chatterjee, C. Federmann, M. Fishel, Y. Graham, B. Haddow, M. Huck, A. J. Yepes, P. Koehn, A. Martins, C. Monz, M. Negri, A. Névéol, M. Neves, M. Post, M. Turchi, & K. Verspoor (Eds.), Proceedings of the Fourth Conference on Machine Translation (Volume 3: Shared Task Papers, Day 2) (pp. 261–266). Association for Computational Linguistics. https://doi.org/10.18653/v1/W19-5435
- DGT-Translation Memory—European Commission. (n.d.). Retrieved November 4, 2024, from https://joint-research-centre.ec.europa.eu/language-technology-resources/dgt-translation-memory_en
- Eisele, A., & Chen, Y. (2010). MultiUN: A Multilingual Corpus from United Nation Documents. In N. Calzolari, K. Choukri, B. Maegaard, J. Mariani, J. Odijk, S. Piperidis, M. Rosner, & D. Tapias (Eds.), Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC’10). European Language Resources Association (ELRA). http://www.lrec-conf.org/proceedings/lrec2010/pdf/686_Paper.pdf
- El-Kishky, A., Chaudhary, V., Guzmán, F., & Koehn, P. (2020). CCAligned: A Massive Collection of Cross-Lingual Web-Document Pairs. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 5960–5969. https://doi.org/10.18653/v1/2020.emnlp-main.480
- El-Kishky, A., Renduchintala, A., Cross, J., Guzmán, F., & Koehn, P. (2021). XLEnt: Mining a Large Cross-lingual Entity Dataset with Lexical-Semantic-Phonetic Word Alignment. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 10424–10430. https://doi.org/10.18653/v1/2021.emnlp-main.814
- Fan, A., Bhosale, S., Schwenk, H., Ma, Z., El-Kishky, A., Goyal, S., Baines, M., Celebi, O., Wenzek, G., Chaudhary, V., Goyal, N., Birch, T., Liptchinsky, V., Edunov, S., Grave, E., Auli, M., & Joulin, A. (2020). Beyond English-Centric Multilingual Machine Translation (No. arXiv:2010.11125). arXiv. https://doi.org/10.48550/arXiv.2010.11125
- García-Martínez, M., Bié, L., Cerdà, A., Estela, A., Herranz, M., Krišlauks, R., Melero, M., O’Dowd, T., O’Gorman, S., Pinnis, M., Stafanovič, A., Superbo, R., & Vasiļevskis, A. (2021). Neural Translation for European Union (NTEU). 316–334. https://aclanthology.org/2021.mtsummit-up.23
- Gibert, O. de, Nail, G., Arefyev, N., Bañón, M., Linde, J. van der, Ji, S., Zaragoza-Bernabeu, J., Aulamo, M., Ramírez-Sánchez, G., Kutuzov, A., Pyysalo, S., Oepen, S., & Tiedemann, J. (2024). A New Massive Multilingual Dataset for High-Performance Language Technologies (No. arXiv:2403.14009). arXiv. http://arxiv.org/abs/2403.14009
- Koehn, P. (2005). Europarl: A Parallel Corpus for Statistical Machine Translation. Proceedings of Machine Translation Summit X: Papers, 79–86. https://aclanthology.org/2005.mtsummit-papers.11
- Kreutzer, J., Caswell, I., Wang, L., Wahab, A., Van Esch, D., Ulzii-Orshikh, N., Tapo, A., Subramani, N., Sokolov, A., Sikasote, C., Setyawan, M., Sarin, S., Samb, S., Sagot, B., Rivera, C., Rios, A., Papadimitriou, I., Osei, S., Suarez, P. O., … Adeyemi, M. (2022). Quality at a Glance: An Audit of Web-Crawled Multilingual Datasets. Transactions of the Association for Computational Linguistics, 10, 50–72. https://doi.org/10.1162/tacl_a_00447
- Rozis, R.,Skadiņš, R (2017). Tilde MODEL - Multilingual Open Data for EU Languages. https://aclanthology.org/W17-0235
- Schwenk, H., Chaudhary, V., Sun, S., Gong, H., & Guzmán, F. (2019). WikiMatrix: Mining 135M Parallel Sentences in 1620 Language Pairs from Wikipedia (No. arXiv:1907.05791). arXiv. https://doi.org/10.48550/arXiv.1907.05791
- Schwenk, H., Wenzek, G., Edunov, S., Grave, E., & Joulin, A. (2020). CCMatrix: Mining Billions of High-Quality Parallel Sentences on the WEB (No. arXiv:1911.04944). arXiv. https://doi.org/10.48550/arXiv.1911.04944
- Steinberger, R., Pouliquen, B., Widiger, A., Ignat, C., Erjavec, T., Tufiş, D., & Varga, D. (n.d.). The JRC-Acquis: A Multilingual Aligned Parallel Corpus with 20+ Languages. http://www.lrec-conf.org/proceedings/lrec2006/pdf/340_pdf
- Subramani, N., Luccioni, S., Dodge, J., & Mitchell, M. (2023). Detecting Personal Information in Training Corpora: An Analysis. In A. Ovalle, K.-W. Chang, N. Mehrabi, Y. Pruksachatkun, A. Galystan, J. Dhamala, A. Verma, T. Cao, A. Kumar, & R. Gupta (Eds.), Proceedings of the 3rd Workshop on Trustworthy Natural Language Processing (TrustNLP 2023) (pp. 208–220). Association for Computational Linguistics. https://doi.org/10.18653/v1/2023.trustnlp-1.18
- Tiedemann, J. (23-25). Parallel Data, Tools and Interfaces in OPUS. In N. C. (Conference Chair), K. Choukri, T. Declerck, M. U. Doğan, B. Maegaard, J. Mariani, A. Moreno, J. Odijk, & S. Piperidis (Eds.), Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC’12). European Language Resources Association (ELRA). http://www.lrec-conf.org/proceedings/lrec2012/pdf/463_Paper
- Ziemski, M., Junczys-Dowmunt, M., & Pouliquen, B. (n.d.). The United Nations Parallel Corpus v1.0. https://aclanthology.org/L16-1561



</details>


### Instruction Tuning Data

This model has been fine-tuned on ~404k instructions, primarily targeting machine translation performance for Catalan, English, and Spanish. Additional instruction data for other European and closely related Iberian languages was also included, as it yielded a positive impact on the languages of interest. That said, the performance in these additional languages is not guaranteed due to the limited amount of available data and the lack of resources for thorough testing.

Tasks related to machine translation are included, but no chat data was used in the fine-tuning process.

Click the expand button below to see the full list of tasks included in the finetuning data.

<details>
<summary>Data Sources</summary>



| Task                             | Source                                                                                   | Languages                                                      | Count  |
|----------------------------------|------------------------------------------------------------------------------------------|----------------------------------------------------------------|--------|
| Chat                             | N/A                                                                                      | N/A                                                            | 0      |
| Multi-reference Translation      | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | mixed                                                          | 10000  |
| Paraphrase 		           | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | mixed                                                          | 3521   |
| Named-entity Recognition         | [AnCora-Ca-NER](https://huggingface.co/datasets/projecte-aina/ancora-ca-ner)                | ca                                                             | 12059  |
| Named-entity Recognition         | [BasqueGLUE](https://huggingface.co/datasets/orai-nlp/basqueGLUE), [EusIE](https://huggingface.co/datasets/HiTZ/EusIE) | eu                                                             | 4304   |
| Named-entity Recognition         | [SLI NERC Galician Gold Corpus](https://github.com/xavier-gz/SLI_Galician_Corpora)          | gl                                                             | 6483   |
| Named-entity Recognition         | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | pt                                                             | 854    |
| Named-entity Recognition         | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | nl                                                             | 800    |
| Named-entity Recognition         | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | es                                                             | 1654   |
| Named-entity Recognition         | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en                                                             | 1671   |
| Named-entity Recognition         | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | ru                                                             | 800    |
| Named-entity Recognition         | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | it                                                             | 858    |
| Named-entity Recognition         | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | fr                                                             | 857    |
| Named-entity Recognition         | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | de                                                             | 1312   |
| Terminology-aware Translation    | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-ru                                                          | 50     |
| Terminology-aware Translation    | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-fr                                                          | 29     |
| Automatic Post Edition           | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-fr                                                          | 6133   |
| Automatic Post Edition           | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-nl                                                          | 9077   |
| Automatic Post Edition           | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-pt                                                          | 5762   |
| Automatic Post Edition           | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | de-en                                                          | 10000  |
| Automatic Post Edition           | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-de                                                          | 10000  |
| Machine Translation Evaluation   | TowerBlocks-sample                                                                       | en-ru, en-pl, ru-en, en-de, en-ru, de-fr, de-en, en-de             | 353    |
| Machine Translation Evaluation   | BSC                                                                                      | four pivot languages (eu, es, ca, gl) paired with European languages (bg, cs, da, de, el, en, et, fi, fr, ga, hr, hu, it, lt, lv, mt, nl, pl, pt, ro, sk, sl, sv)                                   | 9700   |
| General Machine Translation      | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | nl-en, en-ru, it-en, fr-en, es-en, en-fr, ru-en, fr-de, en-nl, de-fr  | 500    |
| General Machine Translation      | BSC                                                                                      | three pivot languages (es, ca, en) paired with European languages (ast, arn, arg, bg, cs, cy, da, de, el, et, fi, ga, gl, hr, it, lt, lv, mt, nb, nn, nl, oc, pl, pt, ro, ru, sk, sl, sr, sv, uk, eu)                                    | 9350   |
| Fill-in-the-Blank                | BSC                                                                                      | European languages (cs, da, de, el, et, fi, fr, ga, hr, hu, it, lt, lv, mt, nl, pl, pt, ro, sk, sl, sv), pivot in ca, es, eu, gl, en                                     | 11500  |
| Document-level Translation       | BSC                                                                                      | two pivot languages (es, en) paired with European languages (bg, cs, da, de, el, et, fi, fr, hu, it, lt, lv, nl, pl, pt, ro, ru, sk, sv)                                                             | 7600   |
| Paragraph-level Translation      | BSC                                                                                      | two pivot languages (es, en) paired with European languages (bg, cs, da, de, el, et, fi, fr, hu, it, lt, lv, nl, pl, pt, ro, ru, sk, sv)                                                              | 7600   |
| Contextual Machine Translation   | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-it                                                          | 348    |
| Contextual Machine Translation   | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-ru                                                          | 454    |
| Contextual Machine Translation   | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-fr                                                          | 369    |
| Contextual Machine Translation   | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-nl                                                          | 417    |
| Contextual Machine Translation   | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-es                                                          | 431    |
| Contextual Machine Translation   | [TowerBlocks](https://huggingface.co/datasets/Unbabel/TowerBlocks-v0.2)                     | en-de                                                          | 558    |

 



</details>

## Evaluation

Below are the evaluation results on the Flores+200 devtest set, compared against the state-of-the-art MADLAD400-7B model ([Kudugunta, S., et al.](https://arxiv.org/abs/2309.04662)). These results cover translation directions between CA-XX, ES-XX, EN-XX, as well as XX-CA, XX-ES, and XX-EN. The metrics have been computed excluding Asturian, Aranese, and Aragonese as we report them separately. The evaluation was conducted using [MT Lens](https://github.com/langtech-bsc/mt-evaluation) following the standard setting (beam search with beam size 5, limiting the translation length to 500 tokens). We report the following metrics:

<details>
<summary>Click to show metrics details</summary>

- `BLEU`: Sacrebleu implementation. Signature: nrefs:1— case:mixed— eff:no— tok:13a— smooth:exp—version:2.3.1
- `TER`: Sacrebleu implementation.
- `ChrF`: Sacrebleu implementation.
- `Comet`: Model checkpoint: "Unbabel/wmt22-comet-da".
- `Comet-kiwi`: Model checkpoint: "Unbabel/wmt22-cometkiwi-da".
- `Bleurt`: Model checkpoint: "lucadiliello/BLEURT-20".
- `MetricX`: Model checkpoint: "google/metricx-23-xl-v2p0".
- `MetricX-QE`: Model checkpoint: "google/metricx-23-qe-xl-v2p0".

</details>

### English

This section presents the evaluation metrics for the English translation tasks.


|                       |   Bleu↑ |   Ter↓ |   ChrF↑ |   Comet↑ |   Comet-kiwi↑ |   Bleurt↑ |   MetricX↓ |   MetricX-QE↓ |
|:---------------------------------|-------:|------:|-------:|--------:|-------------:|---------:|----------:|-------------:|
| **EN-XX** | | | | | | | | |
| SalamandraTA-7b-instruct |  **36.29** | **50.62** |  63.3  |    **0.89** |         **0.85** |     **0.79** |      **1.02** |         **0.94** |
| MADLAD400-7B                     |  35.73 | 51.87 |  **63.46** |    0.88 |         **0.85** |     **0.79** |      1.16 |         1.1  |
| SalamandraTA-7b-base                   |  34.99 | 52.64 |  62.58 |    0.87 |         0.84 |     0.77 |      1.45 |         1.23 |
| **XX-EN** | | | | | | | | |
| SalamandraTA-7b-instruct |  **44.69** | **41.72** |  68.17 |    **0.89** |         0.85 |      **0.8** |      **1.09** |         **1.11** |
| SalamandraTA-7b-base                   |  44.12 | 43    |  **68.43** |    **0.89** |         0.85 |      **0.8** |      1.13 |         1.22 |
| MADLAD400-7B                     |  43.2  | 43.33 |  67.98 |    **0.89** |         **0.86** |      0.8 |      1.13 |         1.15 |


<img src="./images/bleu_en.png" alt="English" width="100%"/>

### Spanish

This section presents the evaluation metrics for the Spanish translation tasks.


|                       |   Bleu↑ |   Ter↓ |   ChrF↑ |   Comet↑ |   Comet-kiwi↑ |   Bleurt↑ |   MetricX↓ |   MetricX-QE↓ |
|:---------------------------------|-------:|------:|-------:|--------:|-------------:|---------:|----------:|-------------:|
| **ES-XX** | | | | | | | | |
| SalamandraTA-7b-instruct |  **23.67** | **65.71** |  53.55 |    **0.87** |         0.82 |     **0.75** |      **1.04** |         **1.05** |
| MADLAD400-7B                     |  22.48 | 68.91 |  **53.93** |    0.86 |         **0.83** |     **0.75** |      1.09 |         1.14 |
| SalamandraTA-7b-base                   |  21.63 | 70.08 |  52.98 |    0.86 |         **0.83** |     0.74 |      1.24 |         1.12 |
| **XX-ES** | | | | | | | | |
| SalamandraTA-7b-instruct |  **25.56** | **62.51** |  52.69 |    **0.85** |         0.83 |     0.73 |      **0.94** |         **1.33** |
| MADLAD400-7B                   |  24.85 | 61.82 |  **53**    |    **0.85** |         **0.84** |     **0.74** |      1.05 |         1.5  |
| SalamandraTA-7b-base                     |  24.71 | 62.33 |  52.96 |    **0.85** |         **0.84** |     0.73 |      1.06 |         1.37 |


<img src="./images/bleu_es.png" alt="English" width="100%"/>



### Catalan

This section presents the evaluation metrics for the Catalan translation tasks.


|                       |   Bleu↑ |   Ter↓ |   ChrF↑ |   Comet↑ |   Comet-kiwi↑ |   Bleurt↑ |   MetricX↓ |   MetricX-QE↓ |
|:---------------------------------|-------:|------:|-------:|--------:|-------------:|---------:|----------:|-------------:|
| **CA-XX** | | | | | | | | |
| MADLAD400-7B                  |  **29.37** | 59.01 |  **58.47** |    **0.87** |         **0.81** |     **0.77** |      **1.08** |         1.31 |
| SalamandraTA-7b-instruct      |  29.23 | **58.32** |  57.76 |    **0.87** |         **0.81** |     **0.77** |      **1.08** |         **1.22** |
| SalamandraTA-7b-base          |  29.06 | 59.32 |  58    |    **0.87** |         **0.81** |     0.76 |      1.23 |         1.28 | 
| **XX-CA** | | | | | | | | |
| SalamandraTA-7b-instruct |  **33.64** | **54.49** |  59.03     |    **0.86** |         0.8      |     **0.75** |      **1.07** |         **1.6**  |
| MADLAD400-7B             |  33.02     | 55.01     |  59.38     |    **0.86** |         **0.81** |     **0.75** |      1.18 |         1.79 |
| SalamandraTA-7b-base     |  32.75     | 55.78     |  **59.42** |    **0.86** |         **0.81** |     **0.75** |      1.17 |         1.63 |


<img src="./images/bleu_ca.png" alt="English" width="100%"/>


### Low-Resource Languages of Spain

The tables below summarize the performance metrics for English, Spanish, and Catalan to Asturian, Aranese and Aragonese.

#### English-XX

|                        | Source   | Target   |   Bleu↑ |    Ter↓ |   ChrF↑ |
|:---------------------------------|:---------|:---------|-------:|-------:|-------:|
| SalamandraTA-7b-instruct               | en       | ast      |  **31.49** |  **54.01** |  **60.65** |
| SalamandraTA-7b-base                   | en       | ast      |  26.4      |  64.02 |  57.35 |
| nllb-3.3B                              | en       | ast      |  22.02     |  77.26 |  51.4  |
| | | | | | |
| SalamandraTA-7b-instruct | en       | arn      |  **13.04** |  **87.13** |  **37.56** |
| SalamandraTA-7b-base                   | en       | arn      |   8.36 |  90.85 |  34.06 |
| | | | | | |
| SalamandraTA-7b-instruct | en       | arg      |  **20.43** |  **65.62** |  **50.79** |
| SalamandraTA-7b-base                   | en       | arg      |  12.24 |  73.48 |  44.75 |


#### Spanish-XX

|                        | Source   | Target   |   Bleu↑ |    Ter↓ |   ChrF↑ |
|:---------------------------------|:---------|:---------|-------:|-------:|-------:|
| SalamandraTA-7b-instruct         | es       | ast      |  **21.28** |  **68.11** |  **52.73** |
| SalamandraTA-7b-base             | es       | ast      |  17.65 |  75.78 |  51.05 |
| salamandraTA2B                   | es       | ast      |  16.68 |  77.29 |  49.46 |
| nllb-3.3B                        | es       | ast      |  11.85 | 100.86 |  40.27 |
| | | | | | |
| SalamandraTA-7b-base             | es       | arn      |  **29.19** |  **71.85** |  **49.42** |
| SalamandraTA-7b-instruct         | es       | arn      |  26.82 |  74.04 |  47.55 |
| salamandraTA2B                   | es       | arn      |  25.41 |  74.71 |  47.33 |
| | | | | | |
| SalamandraTA-7b-base             | es       | arg      |  **53.96** |  **31.51** |  **76.08** |
| SalamandraTA-7b-instruct         | es       | arg      |  47.54 |  36.57 |  72.38 |
| salamandraTA2B                   | es       | arg      |  44.57 |  37.93 |  71.32 |


#### Catalan-XX


|                        | Source   | Target   |   Bleu↑ |    Ter↓ |   ChrF↑ |
|:---------------------------------|:---------|:---------|-------:|-------:|-------:|
| SalamandraTA-7b-instruct | ca       | ast      |  **27.86** |  **58.19** |  57.98 |
| SalamandraTA-7b-base                   | ca       | ast      |  26.11 |  63.63 |  **58.08** |
| salamandraTA2B                   | ca       | ast      |  25.32 |  62.59 |  55.98 |
| nllb-3.3B                        | ca       | ast      |  17.17 |  91.47 |  45.83 |
| | | | | | |
| SalamandraTA-7b-base                   | ca       | arn      |  **17.77** |  **80.88** |  **42.12** |
| SalamandraTA-7b-instruct | ca       | arn      |  16.45 |  82.01 |  41.04 |
| salamandraTA2B                   | ca       | arn      |  15.37 |  82.76 |  40.53 |
| | | | | | |
| SalamandraTA-7b-base                   | ca       | arg      |  **22.53** |  **62.37** |  **54.32** |
| SalamandraTA-7b-instruct | ca       | arg      |  21.62 |  63.38 |  53.01 |
| salamandraTA2B                   | ca       | arg      |  18.6  |  65.82 |  51.21 |



## Ethical Considerations and Limitations

Detailed information on the work done to examine the presence of unwanted social and cognitive biases in the base model can be found 
at [Salamandra-7B model card](https://huggingface.co/BSC-LT/salamandra-7b).
With regard to MT models, no specific analysis has yet been carried out in order to evaluate potential biases or limitations in translation 
accuracy across different languages, dialects, or domains. However, we recognize the importance of identifying and addressing any harmful stereotypes, 
cultural inaccuracies, or systematic performance discrepancies that may arise in Machine Translation. As such, we plan to perform more analyses as soon 
as we have implemented the necessary metrics and methods within our evaluation framework [MT Lens](https://github.com/langtech-bsc/mt-evaluation).
Note that the model has only undergone preliminary instruction tuning. We urge developers to consider potential limitations and conduct safety testing and tuning tailored to their specific applications.

## Additional information

### Author
The Language Technologies Unit from Barcelona Supercomputing Center.

### Contact
For further information, please send an email to <langtech@bsc.es>.

### Copyright
Copyright(c) 2024 by Language Technologies Unit, Barcelona Supercomputing Center.

### Funding
This work has been promoted and financed by the Government of Catalonia through the [Aina Project](https://projecteaina.cat/).

This work is funded by the _Ministerio para la Transformación Digital y de la Función Pública_ - Funded by EU – NextGenerationEU 
within the framework of [ILENIA Project](https://proyectoilenia.es/) with reference 2022/TL22/00215337.



### Disclaimer
Be aware that the model may contain biases or other unintended distortions. 
When third parties deploy systems or provide services based on this model, or use the model themselves, 
they bear the responsibility for mitigating any associated risks and ensuring compliance with applicable regulations, 
including those governing the use of Artificial Intelligence.

The Barcelona Supercomputing Center, as the owner and creator of the model, shall not be held liable for any outcomes resulting from third-party use.

### License
[Apache License, Version 2.0](https://www.apache.org/licenses/LICENSE-2.0)