Spaces:
Running
on
Zero
Running
on
Zero
da03
commited on
Commit
·
9428a07
1
Parent(s):
e2618b3
app.py
CHANGED
|
@@ -1,5 +1,4 @@
|
|
| 1 |
import spaces
|
| 2 |
-
import torch
|
| 3 |
import gradio as gr
|
| 4 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
| 5 |
|
|
@@ -12,14 +11,19 @@ def preprocess(num):
|
|
| 12 |
reversed_num = ' '.join(num[::-1])
|
| 13 |
return reversed_num
|
| 14 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
@spaces.GPU
|
| 16 |
def predict_product(num1, num2):
|
| 17 |
input_text = f'{preprocess(num1)} * {preprocess(num2)} ='
|
| 18 |
inputs = tokenizer(input_text, return_tensors='pt').to('cuda' if torch.cuda.is_available() else 'cpu')
|
| 19 |
model.to('cuda' if torch.cuda.is_available() else 'cpu')
|
| 20 |
outputs = model.generate(**inputs, max_new_tokens=40)
|
| 21 |
-
|
| 22 |
-
|
|
|
|
| 23 |
return input_text, raw_output, prediction
|
| 24 |
|
| 25 |
demo = gr.Interface(
|
|
@@ -31,7 +35,13 @@ demo = gr.Interface(
|
|
| 31 |
gr.Textbox(label='Predicted Product')
|
| 32 |
],
|
| 33 |
title='GPT-2 Multiplication Predictor',
|
| 34 |
-
description='Enter two numbers up to 9 digits each and get the predicted product.'
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
)
|
| 36 |
|
| 37 |
demo.launch()
|
|
|
|
| 1 |
import spaces
|
|
|
|
| 2 |
import gradio as gr
|
| 3 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
| 4 |
|
|
|
|
| 11 |
reversed_num = ' '.join(num[::-1])
|
| 12 |
return reversed_num
|
| 13 |
|
| 14 |
+
def postprocess(raw_output):
|
| 15 |
+
prediction = raw_output.replace(' ', '')[::-1]
|
| 16 |
+
return prediction
|
| 17 |
+
|
| 18 |
@spaces.GPU
|
| 19 |
def predict_product(num1, num2):
|
| 20 |
input_text = f'{preprocess(num1)} * {preprocess(num2)} ='
|
| 21 |
inputs = tokenizer(input_text, return_tensors='pt').to('cuda' if torch.cuda.is_available() else 'cpu')
|
| 22 |
model.to('cuda' if torch.cuda.is_available() else 'cpu')
|
| 23 |
outputs = model.generate(**inputs, max_new_tokens=40)
|
| 24 |
+
output = outputs[0][inputs['input_ids'].shape[-1]:]
|
| 25 |
+
raw_output = tokenizer.decode(output, skip_special_tokens=True)
|
| 26 |
+
prediction = postprocess(raw_output)
|
| 27 |
return input_text, raw_output, prediction
|
| 28 |
|
| 29 |
demo = gr.Interface(
|
|
|
|
| 35 |
gr.Textbox(label='Predicted Product')
|
| 36 |
],
|
| 37 |
title='GPT-2 Multiplication Predictor',
|
| 38 |
+
description='Enter two numbers up to 9 digits each and get the predicted product.',
|
| 39 |
+
article="""
|
| 40 |
+
### Additional Resources
|
| 41 |
+
- [Paper: From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step](https://arxiv.org/pdf/2405.14838)
|
| 42 |
+
- [Code Repository](https://github.com/da03/Internalize_CoT_Step_by_Step)
|
| 43 |
+
- [Tweet Announcement](https://twitter.com/yuntiandeng/status/1795854740879774036)
|
| 44 |
+
"""
|
| 45 |
)
|
| 46 |
|
| 47 |
demo.launch()
|