doyoungkim commited on
Commit
4a08d86
1 Parent(s): 9b42e78

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +9 -8
README.md CHANGED
@@ -13,16 +13,17 @@ Our overall explanation models along with ablations can be found in our [paper](
13
  |-|-|
14
  |[Flipped_11B](https://huggingface.co/seonghyeonye/flipped_11B)|11 billion|
15
  |[Flipped_3B](https://huggingface.co/seonghyeonye/flipped_3B)|3 billion|
16
- Here is how to use the model in PyTorch:
 
17
  ```python
18
- from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
19
- tokenizer = AutoTokenizer.from_pretrained("seonghyeonye/flipped_3B")
20
- model = AutoModelForSeq2SeqLM.from_pretrained("seonghyeonye/flipped_3B")
21
- inputs = tokenizer.encode("input: this is the best cast iron skillet you will ever buy\noutput: Positive", return_tensors="pt")
22
- outputs = model.generate(inputs)
23
- print(tokenizer.decode(outputs[0]))
24
  ```
25
- If you want to use another checkpoint, please replace the path in `AutoTokenizer` and `AutoModelForSeq2SeqLM`.
 
26
  **Note: the model was trained with fp32 activations. As such, we highly discourage running inference with fp16.**
27
 
28
  # Training procedure
 
13
  |-|-|
14
  |[Flipped_11B](https://huggingface.co/seonghyeonye/flipped_11B)|11 billion|
15
  |[Flipped_3B](https://huggingface.co/seonghyeonye/flipped_3B)|3 billion|
16
+ Here is how to download the model in PyTorch:
17
+
18
  ```python
19
+ import torch
20
+ from transformers import T5Tokenizer, T5ForConditionalGeneration
21
+
22
+ model = T5ForConditionalGeneration.from_pretrained("seonghyeonye/flipped_11B")
23
+ tokenizer = T5Tokenizer.from_pretrained("seonghyeonye/flipped_11B")
 
24
  ```
25
+ If you want to use another checkpoint, please replace the path in `T5Tokenizer` and `T5ForConditionalGeneration`.
26
+ We also provide a quick [Jupyter Notebook](https://github.com/seonghyeonye/Flipped-Learning/blob/master/flipped_inference.ipynb) where you can inference with our method.
27
  **Note: the model was trained with fp32 activations. As such, we highly discourage running inference with fp16.**
28
 
29
  # Training procedure