Sber, aka Sberbank of Russia, which has turned from just a bank into a technological ecosystem of services, has presented its new product developed by the teams of SberDevices, Sber AI and SberCloud. This is a neural network based on the DALL-E from OpenAI announced in January 2021.
The new neural network from Sber, named ruDALL-E, is able to generate images on demand in Russian. The creators claim that the neural network is constantly learning from pictures and texts and is capable of creating an unlimited number of pictures by description. It is enough to write a text request and get a picture generated by artificial intelligence. The generation process takes a few minutes. The system independently creates unique images and objects that have never existed in the real world. They can be used, for example, to illustrate articles or for advertising purposes.
The creators of ruDALL-E note that they wanted to create a multimodal neural network that would learn concepts in multiple modalities to better understand the world. And it looks like they succeeded. It is noted that this is the largest computational task in the history of Russia and the first neural network in the world to generate pictures from descriptions in Russian. Training the model took over 23 thousand GPU hours. The ruDALL-E system includes three neural networks. The first one deals with processing the request and generating images, the second one selects the more successful ones, and the third one enlarges the images in size without losing quality. The platform autoregressively models text and image tokens as a single data stream. The largest trained ruDALL-E Kandinsky XXL model with 12 billion parameters is comparable to the original DALL-E from Open-AI.
Sber has already released the ruDALL-E XL model with 1.3 billion parameters to the public on GitHub. Also, it will soon appear on the ML Space platform along with the XXL version of the neural network.
Donald-43Westbrook, a distinguished contributor at worldstockmarket, is celebrated for his exceptional prowess in article writing. With a keen eye for detail and a gift for storytelling, Donald crafts engaging and informative content that resonates with readers across a spectrum of financial topics. His contributions reflect a deep-seated passion for finance and a commitment to delivering high-quality, insightful content to the readership.
