ai-forever
commited on
Commit
•
213baf5
1
Parent(s):
0fadc68
Update README.md
Browse files
README.md
CHANGED
@@ -3,3 +3,22 @@
|
|
3 |
**Ru**ssian **D**iffusion **O**n **L**anguage **P**icture **H**yper-modality Transformer
|
4 |
|
5 |
<img src="https://raw.githubusercontent.com/shonenkov/ru-dolph/master/pics/rudolph-generated.png?token=AHV2MCORXON3DROFL7FBBQ3B4CW4S" height="60" border="2"/>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
**Ru**ssian **D**iffusion **O**n **L**anguage **P**icture **H**yper-modality Transformer
|
4 |
|
5 |
<img src="https://raw.githubusercontent.com/shonenkov/ru-dolph/master/pics/rudolph-generated.png?token=AHV2MCORXON3DROFL7FBBQ3B4CW4S" height="60" border="2"/>
|
6 |
+
|
7 |
+
|
8 |
+
Model was trained by [Sber AI](https://github.com/sberbank-ai) and [SberDevices](https://sberdevices.ru/) teams.
|
9 |
+
* Task: `text2image generation`; `self reranking`; `text reranking`; `image reranking`; `image2text generation`; `zero-shot image classification`;
|
10 |
+
* Language: `Russian`
|
11 |
+
* Type: `encoder-decoder`
|
12 |
+
* Num Parameters: `350M`
|
13 |
+
* Training Data Volume: `35 million text-image pairs`
|
14 |
+
|
15 |
+
|
16 |
+
# Model Description
|
17 |
+
|
18 |
+
RuDOLPH 350M is a fast and light text-image-text transformer (350M GPT-3) designed for a quick and easy fine-tuning setup for the solution of various tasks: from generating images by text description and image classification to visual question answering and more. This model demonstrates the power of Hyper-Modal Transformers.
|
19 |
+
|
20 |
+
# Sparse Attention Mask
|
21 |
+
|
22 |
+
The primary proposed method is to modify the sparse transformer's attention mask to better control multi-modalities. It allows us to calculate the transitions of modalities in both directions, unlike another similar work DALL-E Transformer, which used only one direction, "text to image". The proposed "image to right text" direction is achieved by extension sparse attention mask to the right for auto-repressively text generation with image condition without attention to left text.
|
23 |
+
|
24 |
+
<img src="https://raw.githubusercontent.com/shonenkov/ru-dolph/master/pics/attention_masks.png?token=AHV2MCP7BH3CQBAK74UVA7TB4CXQE" height="40" border="2"/>
|