Update README.md
Browse files
README.md
CHANGED
|
@@ -1,18 +1,18 @@
|
|
| 1 |
---
|
| 2 |
-
title:
|
| 3 |
-
emoji:
|
| 4 |
-
colorFrom:
|
| 5 |
-
colorTo:
|
| 6 |
sdk: gradio
|
| 7 |
app_file: app.py
|
| 8 |
pinned: false
|
| 9 |
---
|
| 10 |
|
| 11 |
-
#
|
| 12 |
|
| 13 |
-
|
| 14 |
|
| 15 |
-
### How
|
| 16 |
-
1.
|
| 17 |
-
2.
|
| 18 |
-
3.
|
|
|
|
| 1 |
---
|
| 2 |
+
title: Neural Layer Inspector
|
| 3 |
+
emoji: 🧠
|
| 4 |
+
colorFrom: blue
|
| 5 |
+
colorTo: indigo
|
| 6 |
sdk: gradio
|
| 7 |
app_file: app.py
|
| 8 |
pinned: false
|
| 9 |
---
|
| 10 |
|
| 11 |
+
# Neural Layer Inspector
|
| 12 |
|
| 13 |
+
This is a technical utility designed for Deep Learning engineers to inspect the internal attention mechanisms of Vision Transformers (ViT).
|
| 14 |
|
| 15 |
+
### How it works
|
| 16 |
+
1. **Model**: Uses `google/vit-base-patch16-224`.
|
| 17 |
+
2. **Extraction**: Accesses the `attentions` attribute from the model output.
|
| 18 |
+
3. **Visualization**: Maps the 1D attention sequence back to the 2D image coordinate space using interpolation and normalization.
|