Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+vocab.spm filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,158 @@
+---
+language:
+- eo
+- ca
+tags:
+- machine-translation
+- translation
+- marian
+- esperanto
+- catalan
+- neural-machine-translation
+library_name: marian
+pipeline_tag: translation
+license: apache-2.0
+---
+
+# Esperanto → Catalan MarianMT model
+
+This repository contains a **Marian NMT** model for **Esperanto-to-Catalan** machine translation.
+
+## Overview
+
+This model was trained for the translation direction:
+
+- **Source language:** Esperanto (`eo`)
+- **Target language:** Catalan (`ca`)
+
+It is distributed in **Marian format** and is intended to be used with the **Marian decoder**.
+
+## Important note
+
+This model is **not intended for direct inference through the Hugging Face `transformers` library**.
+
+Use **Marian** for inference instead.
+
+## Repository contents
+
+The repository includes the following files:
+
+- `model.npz.best-chrf.npz` — trained Marian model checkpoint
+- `tiny.decoder.yml` — decoder configuration
+- `vocab.spm` — SentencePiece vocabulary
+
+## Requirements
+
+You need a working installation of **Marian NMT**.
+
+For example, on our system the decoder binary is located at:
+
+```bash
+/scratch/project_2005815/members/degibert/MTM25/marian/build/marian-decoder
+````
+
+## Inference
+
+Run decoding from inside the model directory:
+
+```bash
+marian-decoder \
+  -c tiny.decoder.yml \
+  --input input.epo \
+  --output output.cat \
+  --normalize \
+  -m model.npz.best-chrf.npz \
+  --vocabs vocab.spm vocab.spm \
+  --log decode.log \
+  --devices 0
+```
+
+## Example
+
+Input file `input.epo`:
+
+```text
+Ŝi amas danci.
+```
+
+Output file `output.cat`:
+
+```text
+Li encanta ballar.
+```
+
+## Example helper script
+
+You can also run the model with a small shell script such as:
+
+```bash
+#!/usr/bin/env bash
+set -euo pipefail
+
+MARIAN_BIN="/path/to/marian-decoder"
+MODEL_DIR="$(cd "$(dirname "$0")" && pwd)"
+
+INPUT="${1:-input.epo}"
+OUTPUT="${2:-output.cat}"
+LOG="${3:-decode.log}"
+
+"$MARIAN_BIN" \
+  -c "$MODEL_DIR/tiny.decoder.yml" \
+  --input "$MODEL_DIR/$INPUT" \
+  --output "$MODEL_DIR/$OUTPUT" \
+  --normalize \
+  -m "$MODEL_DIR/model.npz.best-chrf.npz" \
+  --vocabs "$MODEL_DIR/vocab.spm" "$MODEL_DIR/vocab.spm" \
+  --log "$MODEL_DIR/$LOG" \
+  --devices 0
+```
+
+## Intended use
+
+This model is intended for:
+
+* research on low-resource machine translation
+* Esperanto–Catalan translation experiments
+* reproducible Marian-based inference
+
+## Limitations
+
+This is a research model and may have limitations including:
+
+* reduced robustness outside the training domain
+* sensitivity to spelling variation and noisy input
+* lower quality on idiomatic, literary, or highly specialised text
+
+Outputs should be reviewed before use in high-stakes or publication settings.
+
+## Training and evaluation
+
+Add here any details you want to share, for example:
+
+* training corpus or data source
+* preprocessing pipeline
+* tokenisation / SentencePiece setup
+* evaluation sets
+* BLEU / chrF results
+
+Example placeholder text:
+
+This model was trained as part of research on low-resource translation involving Esperanto and Catalan. Evaluation was carried out on held-out test data using standard MT metrics.
+
+## Citation
+
+If you use this model, please cite the associated work.
+
+```bibtex
+@misc{degibert2026eo_ca_marian,
+  title  = {Esperanto to Catalan MarianMT Model},
+  author = {Degibert, [Your full name]},
+  year   = {2026},
+  note   = {Model distributed via Hugging Face}
+}
+```
+
+## Acknowledgements
+
+This model was trained using **Marian NMT**
+

model.npz.best-chrf.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c180e7307a456d968a9c6ed3cbb22707a8c7e676c279cac47e8c9acfb7c6c243
+size 68714917

run_model.sh

@@ -0,0 +1,17 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+MARIAN_BIN=<path-to-marian>
+MODEL_DIR=<path-to-model-dir>
+src_file=<path-to-src-file>
+tgt=<tgt-tag> # Choose between cat, spa, eng
+
+cat "$src_file" | sed "s/^/>>${tgt}<< /"  \
+  "$MARIAN_BIN" \
+  -c "$MODEL_DIR/tiny.decoder.yml" \
+  --output "$MODEL_DIR/test.epo-cat.cat.out" \
+  --normalize \
+  -m "$MODEL_DIR/model.npz.best-chrf.npz" \
+  --vocabs "$MODEL_DIR/vocab.spm" "$MODEL_DIR/vocab.spm" \
+  --log "$MODEL_DIR/test.log" \
+  --devices 0

tiny.decoder.yml

@@ -0,0 +1,5 @@
+beam-size: 1
+mini-batch: 32
+maxi-batch: 100
+maxi-batch-sort: src
+skip-cost: True

vocab.spm

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d561cdf0fc7ad693c1bf1fe21732c6434650623ec69dc712aceb36483587914d
+size 805644