AI
- NSML - Platform components and infrastructure for AI distributed training
In this post, we introduce you to the problems that the NSML platform was trying to solve and the actual components that'll be more closely examined in future posts. We'll also explain how NSML Pods were configured for distributed training.
Sep 16, 2022 - CLOps - The heart of the platform
We take a look at operators and executors, the core components of the CLOps platform, and what considerations we had while implementing them.
Aug 2, 2022 - CLOVA CareCall - Developing a dialogue system using HyperCLOVA
A look into the development of CLOVA CareCall, and how HyperCLOVA can improve your open-domain dialogue system development.
Jul 14, 2022 - CLOps - The beginning of CLOps, an ML serving platform
A look into why and how we developed the CLOps platform. This post serves as an introduction to future posts where we will go into more detail on our journey while developing CLOps.
Jun 7, 2022 - On the effect of pre-training corpora on in-context learning by large-scale language model.
We investigated the effect of the source and size of pre-training corpora on in-context few-shot and zero-shot learning in HyperCLOVA, a Korean AI platform based on GPT-3.
May 3, 2022 - Optimization points for HyperCLOVA services
We would like to share how we have applied multi-batch and tested the caching effect in a multi-turn pattern service to optimize HyperCLOVA-based services.
Mar 10, 2022 - Extending the features of HyperCLOVA API
We would like to share our experience and some examples of how we've implemented the Early Stop, Semantic Search, and P-tuning features for HyperCLOVA-based services.
Feb 24, 2022
