This can be used to paraphrase. I recommend using the code I have attached below. You can generate it without using LogProbs, but you are likely to be best served by manually examining the most likely outputs.

```
from transformers import AutoTokenizer, AutoModelWithLMHead
tokenizer = AutoTokenizer.from_pretrained("gpt2")
model = AutoModelWithLMHead.from_pretrained("BigSalmon/ParaphraseParentheses2.0")
```

Example Prompt:
```
the nba is [mask] [mask] viewership.
the nba is ( facing / witnessing / confronted with / suffering from / grappling with ) ( lost / tanking ) viewership...

ai is certain to [mask] the third industrial revolution.
ai is certain to ( breed / catalyze / inaugurate / catalyze / usher in / call forth / turn loose / lend its name to ) the third industrial revolution.

the modern-day knicks are a disgrace to [mask].
the modern-day knicks are a disgrace to the franchise's ( rich legacy / tradition of excellence / uniquely distinguished record ).

HuggingFace is [mask].
HuggingFace is ( an amazing company /
```

```
import torch
prompt = "Insert Your Prompt Here. It is Best To Have a Few Examples Before Like The Example Prompt Shows."
text = tokenizer.encode(prompt)
myinput, past_key_values = torch.tensor([text]), None
myinput = myinput
myinput= myinput.to(device)
logits, past_key_values = model(myinput, past_key_values = past_key_values, return_dict=False)
logits = logits[0,-1]
probabilities = torch.nn.functional.softmax(logits)
best_logits, best_indices = logits.topk(500)
best_words = [tokenizer.decode([idx.item()]) for idx in best_indices]
text.append(best_indices[0].item())
best_probabilities = probabilities[best_indices].tolist()
words = []
for i in range(500):
    m = ([best_words[i]])
    m = str(m)
    m = m.replace("[' ", "").replace("']", "")
    print(m)
```