Downloads · 30 days
8
14% of all-time downloads
pguerrero-igutierrez/Latxa-Qwen3-8B-Literary-v2-ca-eu
Latxa-Qwen3-8B-Literary-v2-ca-eu is a text generation model from pguerrero-igutierrez. Use it when you need the model to write or continue text. It is set up for peft.
Latxa-Qwen3-8B-Literary-v2-ca-eu is a LoRA adapter for Catalan-Basque literary machine translation. It continues fine-tuning from the project general-domain checkpoint and targets bidirectional literary translation fo…
Downloads · 30 days
8
14% of all-time downloads
All-time downloads
56
Public
Repo size
1.1 GB
Likes
0
Public
Click a slice to open those files.
.pt699 MB · 55%
From the Hugging Face model README
Latxa-Qwen3-8B-Literary-v2-ca-eu is a LoRA adapter for Catalan-Basque literary machine translation. It continues fine-tuning from the project general-domain checkpoint and targets bidirectional literary translation for the ca->eu and eu->ca directions.
HiTZ/Latxa-Qwen3-VL-8B-Instructca), Basque (eu)HiTZ/Latxa-Qwen3-VL-8B-Instructpguerrero-igutierrez/Latxa-Qwen3-8B-General-eu-capguerrero-igutierrez/Latxa-Qwen3-8B-Literary-v2-ca-eupguerrero-igutierrez/mt-domain-adaptation-ca-euThis model is intended for research on low-resource Catalan-Basque literary translation, especially in settings where in-domain parallel data is scarce and synthetic back-translation data is used for adaptation.
Supported prompting directions:
eu->ca: Itzuli testu hau euskaratik katalanera:\n\n{source}ca->eu: Tradueix aquest text del català al basc:\n\n{source}This adapter was trained on the same literary corpora as literaryv1, built through Spanish-pivot synthetic data generation and back-translation:
backtranslated-corpus/ca-literary_trilingual.jsonbacktranslated-corpus/eu-literary-EhuHac.jsonlThe model was then continued from the project general checkpoint rather than trained directly from the base model.
q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj5e-5Results on the literary held-out test set:
| Direction | chrF++ | BLEU | TER | COMET |
|---|---|---|---|---|
eu->ca | 34.51 | 7.44 | 87.66 | 68.72 |
ca->eu | 25.87 | 2.31 | 100.74 | 64.34 |
| Overall | 30.02 | 5.17 | 93.81 | 65.44 |
In the project experiments, this continued-adaptation literary model performed slightly below the direct literary SFT model (literaryv1) across the reported literary metrics.
If you use this model, please cite the project repository:
@misc{guerrero-gutierrez-2026-caeu-mt,
title = {Domain Adaptation for Catalan-Basque Machine Translation via Synthetic Data and Continued Fine-Tuning},
author = {Guerrero, Paula and Gutierrez, Iker},
year = {2026},
note = {Unpublished manuscript}
}
[email protected][email protected]