Update README.md
Browse files
README.md
CHANGED
@@ -16,23 +16,35 @@ This model is a fine-tuned model of Roberta-large applied on RACE
|
|
16 |
|
17 |
#### How to use
|
18 |
|
19 |
-
```
|
20 |
-
|
21 |
-
|
22 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
dataset = datasets.load_dataset(
|
24 |
"race",
|
25 |
"all",
|
26 |
split=["train", "validation", "test"],
|
27 |
)training_examples = dataset[0]
|
28 |
evaluation_examples = dataset[1]
|
29 |
-
test_examples = dataset[2]
|
|
|
|
|
|
|
|
|
30 |
context = example["article"]
|
31 |
options = example["options"]
|
32 |
label_example = example["answer"]
|
33 |
-
label_map = {label: i
|
|
|
34 |
choices_inputs = []
|
35 |
-
for ending_idx, (_, ending) in enumerate(
|
|
|
36 |
if question.find("_") != -1:
|
37 |
# fill in the banks questions
|
38 |
question_option = question.replace("_", ending)
|
@@ -51,8 +63,8 @@ label = label_map[label_example]
|
|
51 |
input_ids = [x["input_ids"] for x in choices_inputs]
|
52 |
attention_mask = (
|
53 |
[x["attention_mask"] for x in choices_inputs]
|
54 |
-
# as the senteces follow the same structure,
|
55 |
-
# necessary to check
|
56 |
if "attention_mask" in choices_inputs[0]
|
57 |
else None
|
58 |
)
|
|
|
16 |
|
17 |
#### How to use
|
18 |
|
19 |
+
```python
|
20 |
+
|
21 |
+
import datasets
|
22 |
+
from transformers import RobertaTokenizer
|
23 |
+
from transformers import RobertaForMultipleChoice
|
24 |
+
|
25 |
+
tokenizer = RobertaTokenizer.from_pretrained(
|
26 |
+
'LIAMF-USP/roberta-large-finetuned-race')
|
27 |
+
model = RobertaForMultipleChoice.from_pretrained(
|
28 |
+
"LIAMF-USP/roberta-large-finetuned-race")
|
29 |
dataset = datasets.load_dataset(
|
30 |
"race",
|
31 |
"all",
|
32 |
split=["train", "validation", "test"],
|
33 |
)training_examples = dataset[0]
|
34 |
evaluation_examples = dataset[1]
|
35 |
+
test_examples = dataset[2]
|
36 |
+
|
37 |
+
example=training_examples[0]
|
38 |
+
example_id = example["example_id"]
|
39 |
+
question = example["question"]
|
40 |
context = example["article"]
|
41 |
options = example["options"]
|
42 |
label_example = example["answer"]
|
43 |
+
label_map = {label: i
|
44 |
+
for i, label in enumerate(["A", "B", "C", "D"])}
|
45 |
choices_inputs = []
|
46 |
+
for ending_idx, (_, ending) in enumerate(
|
47 |
+
zip(context, options)):
|
48 |
if question.find("_") != -1:
|
49 |
# fill in the banks questions
|
50 |
question_option = question.replace("_", ending)
|
|
|
63 |
input_ids = [x["input_ids"] for x in choices_inputs]
|
64 |
attention_mask = (
|
65 |
[x["attention_mask"] for x in choices_inputs]
|
66 |
+
# as the senteces follow the same structure,
|
67 |
+
#just one of them is necessary to check
|
68 |
if "attention_mask" in choices_inputs[0]
|
69 |
else None
|
70 |
)
|