Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,3 +1,31 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
---
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
# Perception Encoder Audio-Video
|
| 7 |
+
|
| 8 |
+
## Model Summary
|
| 9 |
+
|
| 10 |
+
Perception Encoder Audio-Video (PE-AV) is a family of state-of-the-art encoders for audio and video understanding trained via scaled contrastive learning, built on top of the [PE image/video encoder](https://arxiv.org/abs/2504.13181) (PE)
|
| 11 |
+
|
| 12 |
+
The model is available in the following sizes:
|
| 13 |
+
|
| 14 |
+
- [`pe-av-small`](https://huggingface.co/facebook/pe-av-small): 12 layers, 209M parameters
|
| 15 |
+
- [`pe-av-base`](https://huggingface.co/facebook/pe-av-base): 16 layers, 396M parameters
|
| 16 |
+
- [`pe-av-large`](https://huggingface.co/facebook/pe-av-large): 28L, 1.597B parameters
|
| 17 |
+
|
| 18 |
+
For each size we additionally provide a version that samples a fixed 16-frames for the video branch for efficiency:
|
| 19 |
+
|
| 20 |
+
- [`pe-av-small-16-frame`](https://huggingface.co/facebook/pe-av-small-16-frame): 12 layers, 209M parameters
|
| 21 |
+
- [`pe-av-base-16-frame`](https://huggingface.co/facebook/pe-av-base-16-frame): 16 layers, 396M parameters
|
| 22 |
+
- [`pe-av-large-16-frame`](https://huggingface.co/facebook/pe-av-large-16-frame): 28L, 1.597B parameters
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
## Usage
|
| 26 |
+
|
| 27 |
+
Install `transformers` starting from version v4.34.0
|
| 28 |
+
|
| 29 |
+
```
|
| 30 |
+
pip install 'transformers>=4.34.0'
|
| 31 |
+
```
|