metadata

license: apache-2.0
base_model:
  - google/flan-t5-small
pipeline_tag: text2text-generation

dafilab/chat-title-generator

Fine-tuned flan-t5-small model for generating short titles from chats.

Model Details

Base model: google/flan-t5-small
Training examples: 10,000
Epochs: 2
Final training loss: 0.778800
Train batch size per device: 4
Total optimization steps: 500

Usage

from transformers import T5ForConditionalGeneration, T5Tokenizer

model = T5ForConditionalGeneration.from_pretrained("dafilab/chat-title-generator")
tokenizer = T5Tokenizer.from_pretrained("dafilab/chat-title-generator", legacy=False)

def generate_chat_title(text):
    input_text = "short title: " + text
    inputs = tokenizer(input_text, return_tensors="pt", truncation=True, max_length=512)
    outputs = model.generate(
        input_ids=inputs.input_ids,
        max_length=64,
        num_beams=4,
        early_stopping=True,
        pad_token_id=tokenizer.pad_token_id,
        eos_token_id=tokenizer.eos_token_id
    )
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

text = """How can I access the GPU of my other computer remotely for ML training?
To access your other computer's GPU remotely for machine learning (ML) training,
you need to set up remote access to the machine and ensure that it can properly leverage the GPU for computations.
There are several ways to do this, depending on your operating system and the tools you prefer to use."""
print(generate_chat_title(text))

Output

Remote Access for Machine Learning