README.md

---
language: 
- es
license: mit
library_name: peft
tags:
- peft
- text2text-generation
- text-generation
base_model: google/mt5-large
---

# PTHQL_language_Limburgish

This is the Spanish (spa_Latn) Phylogenetic Tree Hierarquical QLoRAs (PTHQL) adapter from [Generating from AMRs into High and Low-Resource Languages using Phylogenetic Knowledge and Hierarchical QLoRA Training (HQL)](https://aclanthology.org/2024.inlg-main.7/) used for AMR-to-Text generation.

# Use

This model is the last of 4 hierarquical LoRAs. It is strongly adviseable to load all 4 LoRAs in order.

The following is minimal code to generate Spanish text from an AMR graph:
```
from transformers import MT5ForConditionalGeneration, AutoTokenizer
from peft import PeftModel

model = MT5ForConditionalGeneration.from_pretrained('google/mt5-large')
tokennizer = AutoTokenizer.from_pretrained('google/mt5-large')

model = PeftModel.from_pretrained(model, 'WilliamSotoM/PTHQL_level0_Indo_European')
model = model.merge_and_unload()

model = PeftModel.from_pretrained(model, 'WilliamSotoM/PTHQL_level1_Romance')
model = model.merge_and_unload()

model = PeftModel.from_pretrained(model, 'WilliamSotoM/PTHQL_level2_Iberian_Romance')
model = model.merge_and_unload()

model = PeftModel.from_pretrained(model, 'WilliamSotoM/PTHQL_language_Spanish')
model = model.merge_and_unload()

graph = '''
(w / want-01
   :ARG0 (b / boy)
   :ARG1 (b2 / believe-01
             :ARG0 (g / girl)
             :ARG1 b))
'''
tokenized_input = tokenizer(graph, return_tensors='pt')

with torch.inference_mode():
    prediction = model.generate(**tokenized_input)
    generated_text = tokenizer.batch_decode(prediction, skip_special_tokens=True)[0]

print(f'Generated text:', generated_text)
```

Expected outpu:
```
El chico quiere que la chica le crea.
```