t5-chineseAMR-base-multitask

Model fine-tuned with morph resolution and morph explanation

Example

from transformers import T5Tokenizer, T5ForConditionalGeneration

# Load the tokenizer and model
model_path = "your-finetuned-mengzi-t5-checkpoint"  # Replace with your actual model path or Hugging Face model ID
tokenizer = T5Tokenizer.from_pretrained("Langboat/mengzi-t5-base", legacy=False)
model = T5ForConditionalGeneration.from_pretrained(model_path)

# Input text (format should match the training format)
input_text = "<纠正>小糖人都是可以吃的。"  # correct example input in Chinese
#input_text = "<解释>小糖人都是可以吃的。"  # explan  example input in Chinese

# Tokenize the input
inputs = tokenizer(
    input_text,
    return_tensors="pt",
    max_length=512,
    truncation=True,
    padding=True
)

# Generate output
outputs = model.generate(
    inputs["input_ids"],
    max_length=256,
    num_beams=4,
    early_stopping=True
)
# Decode and print the result
generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
print("Generated output:", generated_text)
# 糖尿病患者都是可以吃的。