Update README.md
Browse files
README.md
CHANGED
@@ -1,7 +1,11 @@
|
|
|
|
|
|
|
|
1 |
from FiDT5 import FiDT5
|
2 |
from transformers import T5Tokenizer
|
3 |
model = FiDT5.from_pretrained('Soyoung97/ListT5-base')
|
4 |
-
texts = [
|
|
|
5 |
"Query: When did Thomas Edison invent the light bulb?, Index: 2, Context: Thomas Edison tried to invent a device for car but failed",
|
6 |
"Query: When did Thomas Edison invent the light bulb?, Index: 3, Context: Coffee is good for diet",
|
7 |
"Query: When did Thomas Edison invent the light bulb?, Index: 4, Context: KEPCO fixes light problems",
|
@@ -9,7 +13,8 @@ texts = ["Query: When did Thomas Edison invent the light bulb?, Index: 1, Contex
|
|
9 |
tok = T5Tokenizer.from_pretrained('t5-base')
|
10 |
raw = tok(texts, return_tensors='pt', padding='max_length', max_length=128, truncation=True)
|
11 |
input_tensors = {'input_ids': raw['input_ids'].unsqueeze(0), 'attention_mask': raw['attention_mask'].unsqueeze(0)}
|
12 |
-
output = model.generate(**input_tensors, max_length=
|
13 |
output_text = tok.batch_decode(output.sequences, skip_special_tokens=True)
|
14 |
output_text
|
15 |
-
>>> [3 1 4 2 5]
|
|
|
|
1 |
+
How to use
|
2 |
+
|
3 |
+
```
|
4 |
from FiDT5 import FiDT5
|
5 |
from transformers import T5Tokenizer
|
6 |
model = FiDT5.from_pretrained('Soyoung97/ListT5-base')
|
7 |
+
texts = [
|
8 |
+
"Query: When did Thomas Edison invent the light bulb?, Index: 1, Context: Lightning strike at Seoul National University",
|
9 |
"Query: When did Thomas Edison invent the light bulb?, Index: 2, Context: Thomas Edison tried to invent a device for car but failed",
|
10 |
"Query: When did Thomas Edison invent the light bulb?, Index: 3, Context: Coffee is good for diet",
|
11 |
"Query: When did Thomas Edison invent the light bulb?, Index: 4, Context: KEPCO fixes light problems",
|
|
|
13 |
tok = T5Tokenizer.from_pretrained('t5-base')
|
14 |
raw = tok(texts, return_tensors='pt', padding='max_length', max_length=128, truncation=True)
|
15 |
input_tensors = {'input_ids': raw['input_ids'].unsqueeze(0), 'attention_mask': raw['attention_mask'].unsqueeze(0)}
|
16 |
+
output = model.generate(**input_tensors, max_length=7, return_dict_in_generate=True, output_scores=True)
|
17 |
output_text = tok.batch_decode(output.sequences, skip_special_tokens=True)
|
18 |
output_text
|
19 |
+
>>> [3 1 4 2 5]
|
20 |
+
```
|