Update README.md
Browse files
README.md
CHANGED
|
@@ -1,33 +0,0 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: bsd-3-clause
|
| 3 |
-
---
|
| 4 |
-
Mirror of the base ProGen2-small model (with slightly modified configuration and forward pass) by [Nijkamp, et al.](https://arxiv.org/abs/2206.13517).
|
| 5 |
-
|
| 6 |
-
See also my github [repo](https://github.com/hugohrban/ProGen2-finetuning/tree/main) for an example of finetuning this model.
|
| 7 |
-
|
| 8 |
-
Example usage:
|
| 9 |
-
|
| 10 |
-
```python
|
| 11 |
-
from transformers import AutoModelForCausalLM
|
| 12 |
-
from tokenizers import Tokenizer
|
| 13 |
-
import torch
|
| 14 |
-
import torch.nn.functional as F
|
| 15 |
-
|
| 16 |
-
# load model and tokenizer
|
| 17 |
-
model = AutoModelForCausalLM.from_pretrained("hugohrban/progen2-small", trust_remote_code=True)
|
| 18 |
-
tokenizer = Tokenizer.from_pretrained("hugohrban/progen2-small")
|
| 19 |
-
tokenizer.no_padding()
|
| 20 |
-
|
| 21 |
-
# prepare input
|
| 22 |
-
prompt = "1MEVVIVTGMSGAGK"
|
| 23 |
-
input_ids = torch.tensor(tokenizer.encode(prompt).ids).to(model.device)
|
| 24 |
-
|
| 25 |
-
# forward pass
|
| 26 |
-
logits = model(input_ids).logits
|
| 27 |
-
|
| 28 |
-
# print output probabilities
|
| 29 |
-
next_token_logits = logits[-1, :]
|
| 30 |
-
next_token_probs = F.softmax(next_token_logits, dim=-1)
|
| 31 |
-
for i in range(tokenizer.get_vocab_size(with_added_tokens=False)):
|
| 32 |
-
print(f"{tokenizer.id_to_token(i)}: {100 * next_token_probs[i].item():.2f} %")
|
| 33 |
-
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|