Latest news and updates from DSML Kazakhstan community
Stay up to date with the latest events
Source: arxiv.org/abs/1706.03762
September 3, 18:00. Venue: Jimmy Yamazaki's stream at twitch.tv/yamazakij
Speaker: Jimmy Yamazaki
About the talk: We will examine the paper and its mechanisms in detail, try to understand how and why attention works, dive deep into transformer architecture, write code, and even train models on small datasets. You can ask questions throughout the talk, and the speaker will be happy to answer them.
Participation is free.
Verigram's algorithm received NIST's highest rating in the 1:N Identification category, which matches one face against many. The algorithm can instantly find facial matches in databases containing millions of records.
Verigram's solution was evaluated on a database of 12 million photographs.
NIST, the National Institute of Standards and Technology, is an agency of the U.S. Department of Commerce responsible for developing and maintaining standards across a broad range of technology fields. The NIST Face Recognition Vendor Test (FRVT) is considered the most authoritative benchmark for recognition algorithms.
Last night, we released our first LLaMA-2 fine-tune, which achieved state-of-the-art results on the MMLU and BigBench Hard benchmarks. The model is open source and can be downloaded—and liked—directly from Hugging Face at the link below.
This is an open version of an AI assistant similar to ChatGPT. It can be fine-tuned or run locally for research purposes. A demo is also available at the following link.
A few important details:
What makes our work unique? We focused on careful data cleaning, an essential part of a scientific approach to data analysis. Rather than cleaning the data in a single pass, we kept improving the process iteratively.
One key step toward better results was removing standard responses from larger AI models, such as “Sorry, as an AI assistant I cannot...”, from the dataset. After that, our model achieved the best results among all available 7-billion-parameter models.
The development team includes members of DSML KZ:
More technical details are available in the blog post about the model's release.
Arbuz.kz has launched Arbuz Chef, a ChatGPT-based bot that recommends dishes and recipes. Each recipe comes with a set of products that can be added to the cart. For now, the feature is available through the Friends loyalty program on iOS and the web.
Project initiator Alibek Utyubayev, Analytics Lead at Arbuz.kz, shared these insights:
