|
|
--- |
|
|
license: afl-3.0 |
|
|
language: |
|
|
- en |
|
|
base_model: |
|
|
- google/flan-t5-xl |
|
|
pipeline_tag: text-classification |
|
|
tags: |
|
|
- personality |
|
|
--- |
|
|
|
|
|
|
|
|
|
|
|
## Model Details |
|
|
|
|
|
* **Model Type:** PersonalityClassifier is a fine-tuned model from `google/flan-t5-xl` using annotation data for personality classification. |
|
|
* **Model Date:** PersonalityClassifier was trained in Jan 2024. |
|
|
* **Paper or resources for more information:** [https://arxiv.org/abs/2504.06868](https://arxiv.org/abs/2504.06868) |
|
|
* **Train data:** [https://huggingface.co/datasets/mirlab/personality_120000](https://huggingface.co/datasets/mirlab/personality_120000) |
|
|
## Requirements |
|
|
|
|
|
* `torch==2.1.0` |
|
|
* `transformers==4.29.0` |
|
|
|
|
|
## How to use the model |
|
|
|
|
|
```python |
|
|
import torch |
|
|
from transformers import T5ForConditionalGeneration, AutoTokenizer |
|
|
|
|
|
# Set device to CUDA if available, otherwise use CPU |
|
|
device = torch.device("cuda" if torch.cuda.is_available() else "cpu") |
|
|
|
|
|
# Load model and tokenizer |
|
|
model_name = "mirlab/PersonalityClassifier" |
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
|
model = T5ForConditionalGeneration.from_pretrained(model_name).to(device) |
|
|
|
|
|
# Define model inference function |
|
|
def modelGenerate(input_text, lm, tokenizer): |
|
|
# Tokenize input text and move to device |
|
|
input_ids = tokenizer(input_text, truncation=True, padding=True, return_tensors='pt')['input_ids'].to(device) |
|
|
|
|
|
# Generate text using the model |
|
|
model_output = lm.generate(input_ids) |
|
|
|
|
|
# Decode generated tokens into text |
|
|
model_answer = tokenizer.batch_decode(model_output, skip_special_tokens=True) |
|
|
|
|
|
return model_answer |
|
|
|
|
|
# Example input text |
|
|
# Format: "[Valence] Statement: [Your Statement]. Trait: [Target Trait]" |
|
|
# Target Trait is among ["Openness", "Conscientiousness", "Extraversion", "Agreeableness", "Neuroticism", "Machiavellianism", "Narcissism", "Psychopathy"]. |
|
|
# Valence indicates positive (+) or negative (-) alignment with the trait. |
|
|
|
|
|
input_texts = "[Valence] Statement: I am outgoing. Trait: Extraversion" |
|
|
|
|
|
# Generate output using the model and print |
|
|
output_texts = modelGenerate(input_texts, model, tokenizer) |
|
|
print(output_texts) |
