README.md

---
license: apache-2.0
language:
- en
pipeline_tag: text2text-generation
---

**flan-t5-small-for-classification**

<img src="https://github.com/Knowledgator/unlimited_classifier/raw/main/images/tree.jpeg" style="display: block; margin: auto;" height="720" width="720">

This is an additional fine-tuned [flan-t5-base](https://huggingface.co/google/flan-t5-base) model on many classification datasets. 

The model supports prompt-tuned classification and is suitable for complex classification settings such as resumes classification by criteria.

You can use the model simply generating the text class name or using our [unlimited-classifier](https://github.com/Knowledgator/unlimited_classifier).

The library allows to set constraints on generation and classify text into millions of classes.

### How to use:

To use it with transformers library take a look into the following code snippet:
```python
# pip install accelerate
from transformers import T5Tokenizer, T5ForConditionalGeneration

tokenizer = T5Tokenizer.from_pretrained("knowledgator/flan-t5-base-for-classification")
model = T5ForConditionalGeneration.from_pretrained("knowledgator/flan-t5-base-for-classification", device_map="auto")

input_text = "Define sentiment of the following text: I love to travel and someday I will see the world."
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")

outputs = model.generate(input_ids)
print(tokenizer.decode(outputs[0]))
```

**Using unlimited-classifier**

```python
# pip install unlimited-classifier

from unlimited_classifier import TextClassifier

classifier = TextClassifier(
    labels=[
        'positive',
        'negative',
        'neutral'    
    ],
    model='knowledgator/flan-t5-base-for-classification',
    tokenizer='knowledgator/flan-t5-base-for-classification',
)
output = classifier.invoke(input_text)
print(output)
```