NYC Data Science Academy Alumni Exclusive Event This Thursday 8/17 11 am- 12 pm(EST), please join me for our Live Learning: Live Coding Image Generation Prompt Improvements. As the name suggests, we'll be working from (mostly) scratch to create prompts for a LLM that will incorporate: - prompts from text-to-image models - feedback (this will be fake, but could actually come from A/B testing, holdout group testing, etc.) in order to see if we can get the LLM to make prompts for the text-to-image model that produce "better" output based on the feedback without us needing to manually tweak the resulting prompt too much. I hope that this will give everyone a better idea of how to prompt these models, how LLMs can use in-context learning, and the limitations of this approach. Zoom Link for Live Learning (please message me if you are not able to enter the room): https://lnkd.in/eScKr6kJ Meeting ID: 946 1359 3549 Passcode: 100607
Vivian S. Zhang’s Post
More Relevant Posts
-
NYC Data Science Academy alumni and friends: This Thursday (8/22) at 11:00 am m(EST) will be our weekly Live Learning: Building LLM Powered Systems (part 4 of N). Our focus will be on: - situations where plain LLMs are insufficient for performing a particular task - handling guardrails and routing for black/white listing behavior and conversation paths - using LLMs to accomplish tasks on their own - how various functionalities in commercial products (e.g. ChatGPT's web search) are implemented Last time we started implementing something like ChatGPT's memory system. This time, we will be extending it to dynamically choose which memories should be in context. In doing so, we will: - examine why our existing implementation does not scale - introduce semantic search as a mechanism for choosing relevant memories - use function calling to hand when to search for memories and what to search with There is no way we can cover all of this in one session, so we will continue this topic for a few weeks. In preparation, if you have not done so already, please watch this recording where I go over setting up GPT4All to run open source LLMs on your machine. Please also install Llama 3.1 8B Instruct, as that will be the model I will be demonstrating with. While this is by no means the only way to do this, it will be the method I use during our session. Zoom Link for Live Learning (please message me if you are not able to enter the room): https://lnkd.in/eScKr6kJ Meeting ID: 946 1359 3549 Passcode: 100607
-
-
Great job! Ben Wellington and Claudia Perlich.
Our Deputy Head of Feature Forecasting Ben Wellington and Head of Strategic Data Science for Investment Management Claudia Perlich are featured in Bloomberg on the use of LLMs in investing. Read more of Justina Lee's piece here: https://lnkd.in/gj5WaEaM
Hedge Funds Battle to Turn ChatGPT from Intern to Analyst
bloomberg.com
-
NYC Data Science Academy alumni and friends: Today (8/8/2024) at 11:00am will be our weekly Live Learning: Building LLM Powered Systems (part 2 of N). Our focus will be on: - situations where plain LLMs are insufficient for performing a particular task - handling guardrails and routing for black/white listing behavior and conversation paths - using LLMs to accomplish tasks on their own - how various functionality in commercial products (e.g. ChatGPT's web search) are implemented We will pick up where we left off, having models that talk to each other, and extend that to more practical applications. The code we ended with last time and will start with today is attached. There is no way we can cover all of this in one session, so we will continue this topic for a few weeks. In preparation, if you have not done so already, please watch https://lnkd.in/ecPHkDYn where we go over setting up https://lnkd.in/eA63wUU7 to run open source LLMs on your machine. Please also install Llama 3.1 8B Instruct, as that will be the model I will be demonstrating with. While this is by no means the only way to do this, it will be the method I use during our session. Additionally, if you need an introduction or refresher on function calling, you can watch https://lnkd.in/eNJx9vZW, as we will be putting some of this to use here. Zoom Link for Live Learning (please message me if you are not able to enter the room): https://lnkd.in/eScKr6kJ Meeting ID: 946 1359 3549 Passcode: 100607
-
-
NYC Data Science Academy alumni, great opportunity!
Hello! We’re testing a special beta program designed specifically for job seekers. We've seen many successful job seekers using RocketReach to contact the hiring manager or recruiter for a specific role, and understand how important it is for you to get to the right person, as opposed to being just another resume in the pile. If you are currently job searching and interested in learning more, please reach out to me directly at [email protected]. This program will be free as we try to better understand how to serve you in the best way possible. #JobSeekers #CareerOpportunities #RocketReach #JobSearch #CareerGrowth
Data Scientist @ Datadog
1yHey Vivian! Is there a recording?