ybracke/transnormer-19c-beta-v01

Transnormer 19th century (beta v01)

This model normalizes spelling variants in historical German text to the modern spelling. It is a fine-tuned version of google/byt5-small on a modified version of the DTA EvalCorpus (1780-1901).

Demo Usage

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
from transformers.generation import GenerationConfig

tokenizer = AutoTokenizer.from_pretrained("ybracke/transnormer-19c-beta-v01")
model = AutoModelForSeq2SeqLM.from_pretrained("ybracke/transnormer-19c-beta-v01")

gen_cfg = GenerationConfig.from_model_config(model.config)
gen_cfg.max_new_tokens = 512

sentence = "Der Officier mußte ſich dazu setzen, man trank und ließ ſich’s wohl ſeyn."
inputs = tokenizer(sentence, return_tensors="pt",)
outputs = model.generate(**inputs, generation_config=gen_cfg)
print(tokenizer.batch_decode(outputs, skip_special_tokens=True))
# >>> ['Der Offizier musste sich dazusetzen, man trank und ließ sich es wohl sein.'

Here is how to use this model with the pipeline API:

from transformers import pipeline

transnormer = pipeline('text2text-generation', model='ybracke/transnormer-19c-beta-v01')
sentence = "Der Officier mußte ſich dazu setzen, man trank und ließ ſich’s wohl ſeyn."
print(transnormer(sentence))
# >>> [{'generated_text': 'Der Offizier musste sich dazusetzen, man trank und ließ sich es wohl sein.'}]

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.76

Framework versions

• Transformers 4.31.0
• Pytorch 2.1.0+cu121
• Datasets 2.18.0
• Tokenizers 0.13.3