Deepseek-R1

​The whole world is talking about DeepSeek. It reached #1 on various Apple and Google app stores around the world.

​Join AI Makerspace to hear about the concepts and code of R1; the first-ever Large Reasoning Model (LRM) released from the DeepSeek team.

​We have some high-level questions to answer, including:

• ​What is DeepSeek’s history of models and suite of models today including DeepSeek-V3 (deepseek-chat) and DeepSeek-R1 (deepseek-reasoner)?

• ​Should we be surprised about this release of a Chain of Thought-powered Reasoner?

• ​How does DeepSeek-R1 compare to leading CoT-style models in the industry, including OpenAI’s o1, o3, Qwen QwQ, or Google’s Gemini Flash Thinking?

• ​Can we expect Anthropic, Meta, Mistral, and others to follow suit and release their own reasoning models, now that the game appears to have changed?

​Perhaps more importantly, we want to dig into the paper and the technical details.

• ​Based on what we know about o1, how is DeepSeek R1 different?

• ​What was the role of DeepSeek-R1-Zero along the path to the development of DeepSeek-R1, which was “trained via large-scale Reinforcement Learning (RL) without Supervised Fine Tuning (SFT),” but encountered “challenges such as poor readability and language mixing?” What can we glean from taking a close look at the training template form the paper?

• ​What does it mean that DeepSeek-R1 incorporated multi-stage training and cold-start data before RL?

• ​What role did SFT play during training, and why did DeepSeek-R1 need two different rounds of RL?

• ​How was DeepSeek-R1 used to distill reasoning capabilities down to a set of 800k samples that could be used for fine-tuning Small Language Models in sizes ranging from 1.5B to 70B?

​Of course, expect lots of vibe checking of the reasoning models we can use out of the box.

​It should be a dope event, so join us live to get your questions answered about what’s really going on with DeepSeek-R1!

​📚 You’ll learn:

• ​How DeepSeek R1 was trained, and what it’s capabilities are today compared to other leading reasoning LRMs

• ​How DeepSeek-R1-Zero set the stage for the next generation of model, and where it fell short

• ​How distillation was used to provide the world with a suite of small, performant, reasoning models

• ​What all this means for the future of LRMs and building, shipping, and sharing production LLM and LRM applications!

​🤓 Who should attend the event:

• ​Aspiring AI Engineers & Leaders who want to understand and build with leading LRMs

​Speakers:

• ​Dr. Greg” Loughnane is the Co-Founder & CEO of AI Makerspace, where he is an instructor for their AI Engineering Bootcamp. Since 2021, he has built and led industry-leading Machine Learning education programs.  Previously, he worked as an AI product manager, a university professor teaching AI, an AI consultant and startup advisor, and an ML researcher.  He loves trail running and is based in Dayton, Ohio.

• ​Chris “The Wiz” Alexiuk is the Co-Founder & CTO at AI Makerspace, where he is an instructor for their AI Engineering Bootcamp. During the day, he is also a Developer Advocate at NVIDIA. Previously, he was a Founding Machine Learning Engineer, Data Scientist, and ML curriculum developer and instructor. He’s a YouTube content creator YouTube who’s motto is “Build, build, build!” He loves Dungeons & Dragons and is based in Toronto, Canada.

​Follow AI Makerspace on LinkedIn and YouTube to stay updated about workshops, new courses, and corporate training opportunities.