byT5_ft_generation / README.md
saadamin2k13's picture
Update README.md
6ea462a verified
metadata
language:
  - en
metrics:
  - bleu
  - meteor
  - chrf
  - comet
  - bertscore
library_name: transformers

Model Card for Model ID

This model card lists fine-tuned byT5 model for the task of Text Generation from Meaning Representation (DRS).

Model Details

We worked on a pre-trained byt5-base model and fine-tuned it with the Parallel Meaning Bank dataset (DRS-Text pairs dataset). Furthermore, we enriched the gold_silver flavors of PMB (release 5.0.0) with different augmentation strategies.

Uses

To use the model, follow the code below for a quick response.


from transformers import ByT5Tokenizer, T5ForConditionalGeneration

# Initialize the tokenizer and model
tokenizer = ByT5Tokenizer.from_pretrained('saadamin2k13/byT5_ft_generation', max_length=512)

model = T5ForConditionalGeneration.from_pretrained('saadamin2k13/byT5_ft_generation')

# Example sentence
example = "male.n.02 Name 'Tom' yell.v.01 Agent -1 Time +1 time.n.08 TPR now"

# Tokenize and prepare the input
x = tokenizer(example, return_tensors='pt', padding=True, truncation=True, max_length=512)['input_ids']

# Generate output
output = model.generate(x)

# Decode and print the output text
pred_text = tokenizer.decode(output[0], skip_special_tokens=True, clean_up_tokenization_spaces=False)
print(pred_text)