|
--- |
|
inference: |
|
parameters: |
|
temperature: 0.5 |
|
widget: |
|
text: "A courier received 50 packages yesterday and twice as many today. All of these should be delivered tomorrow. How many packages should be delivered tomorrow?" |
|
--- |
|
|
|
This model was created using GPT-2 as a base, and fine-tuned upon a dataset of elementary school problems requiring logic and reasoning. |
|
Requires Pytorch |
|
|
|
How to use to infer text |
|
```python |
|
|
|
from transformers import AutoTokenizer, AutoModelForCasualLM |
|
import torch |
|
|
|
type = "gpt2-large" |
|
tokenizer = AutoTokenizer.from_pretrained(type) |
|
model = AutoModelForCausalLM.from_pretrained(type) |
|
|
|
model_path = '../model.pt' |
|
|
|
model = torch.load(model_path) |
|
|
|
your_text = "A courier received 50 packages yesterday and twice as many today. All of these should be delivered tomorrow. How many packages should be delivered tomorrow?" |
|
encoded_text = self.tokenizer.encode(your_text, return_tensors='pt') |
|
outputs = model.generate(encoded_text, max_length=64, do_sample=True, temperature=0.5, top_p=1) |
|
outputs = [tokenizer.decode(output) for output in outputs] |
|
``` |
