luffycodes
commited on
Commit
•
d98c6b5
1
Parent(s):
d73f03f
Update README.md
Browse files
README.md
CHANGED
@@ -4,7 +4,20 @@ language:
|
|
4 |
- en
|
5 |
---
|
6 |
## Model weights for Parallel Roberta-Large model ##
|
7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
8 |
|
9 |
If you use this work, please cite:
|
10 |
Investigating the Role of Feed-Forward Networks in Transformers Using Parallel Attention and Feed-Forward Net Design:
|
|
|
4 |
- en
|
5 |
---
|
6 |
## Model weights for Parallel Roberta-Large model ##
|
7 |
+
|
8 |
+
We provide the [weights](https://huggingface.co/luffycodes/Parallel-Roberta-Large) for the parallel attention and feedforward design for Roberta-Large.
|
9 |
+
|
10 |
+
![pfa (1)](https://github.com/luffycodes/Parallel-Transformers-Pytorch/assets/22951144/e5b76b1c-5fb1-4263-a23b-a61742fe12ae)
|
11 |
+
|
12 |
+
## Evaluation results
|
13 |
+
|
14 |
+
When fine-tuned on downstream tasks, this model achieves the following results:
|
15 |
+
|
16 |
+
Glue test results:
|
17 |
+
|
18 |
+
| Task | MNLI | QQP | QNLI | SST-2 | CoLA | STS-B | MRPC | RTE |
|
19 |
+
|:----:|:----:|:----:|:----:|:-----:|:----:|:-----:|:----:|:----:|
|
20 |
+
| | 89.3 | 91.7 | 94.3 | 96.2 | 64.0 | 91.0 | 90.4 | 80.1 |
|
21 |
|
22 |
If you use this work, please cite:
|
23 |
Investigating the Role of Feed-Forward Networks in Transformers Using Parallel Attention and Feed-Forward Net Design:
|