Commit
•
3dbad8a
1
Parent(s):
f4c6802
Update README.md
Browse files
README.md
CHANGED
|
@@ -15,9 +15,12 @@ base_model:
|
|
| 15 |
- google/vit-base-patch16-224-in21k
|
| 16 |
---
|
| 17 |
|
|
|
|
| 18 |
|
| 19 |
-
|
|
|
|
| 20 |
|
|
|
|
| 21 |
|
| 22 |
- eval_loss: 0.19939416646957397
|
| 23 |
- eval_rouge1: 43.006
|
|
|
|
| 15 |
- google/vit-base-patch16-224-in21k
|
| 16 |
---
|
| 17 |
|
| 18 |
+
This model is a variation of https://huggingface.co/nlpconnect/vit-gpt2-image-captioning
|
| 19 |
|
| 20 |
+
=> Read the blog post here https://ziade.org/2024/03/17/distilvit-image-captioning-model
|
| 21 |
+
=> The training code is here: https://github.com/tarekziade/distilvit
|
| 22 |
|
| 23 |
+
Results after after 3 epochs (and ~45 hours of training)
|
| 24 |
|
| 25 |
- eval_loss: 0.19939416646957397
|
| 26 |
- eval_rouge1: 43.006
|