After the single box ran, we scaled it to 2 and 16 systems with linear scaling. There were no model & code changes required.
Megatron
DeepSpeed
Sharding The model lives on a single block of memory. No need to break it up. The whole thing trains like one giant GPU.
SOFTWARE
-
Linear Scaling AI Model Training Across Multiple Systems
By
–
-
ChatGPT integrated into Apple Intelligence and Siri: Apple-OpenAI strategy
By
–
#ChatGPT arrives in #AppleIntelligence and #Siri → https://youtu.be/Ks1sbFyWrNc In this video, we discuss the features and especially the (brilliant) strategy of @Apple with @OpenAI and #AI
-
DuckDB 1.0: High-Speed Analytics Engine for AI Workflows
By
–
#DuckDB Redefining Analytics for #AI 100x faster Parquet queries vs. Pandas Process billions of rows in seconds In-process #SQL <10MB RAM for setup Integrated #Python/#R for #ML workflows Version 1.0 stability + @MotherDuckHQ backing #DataAnalytics #TechTrends
-
Disable Agent Messages in FlowiseAI ChatEmbed Configuration
By
–
You can turn it off via showAgentMessages – https://
github.com/FlowiseAI/Flow
iseChatEmbed?tab=readme-ov-file#configuration
… -
IBM Watsonx Integration Guide for Flowise Platform
By
–
To use IBM Watsonx within Flowise:
— FlowiseAI (@FlowiseAI) 11 décembre 2024
1. Create an account on Watsonx
2. Create a project, and select the model
3. Create API Key on IBM Cloud Console
4. Grab the URL, Project ID, and Model Name
Step by step guide: https://t.co/Q6VOqxzcv7
Thanks @_Eduard26 for contributions! pic.twitter.com/U8IsrHLm7kTo use IBM Watsonx within Flowise: 1. Create an account on Watsonx
2. Create a project, and select the model
3. Create API Key on IBM Cloud Console
4. Grab the URL, Project ID, and Model Name Step by step guide: https://
docs.flowiseai.com/integrations/l
angchain/chat-models/ibm-watsonx
… Thanks @_Eduard26 for contributions! -
IBM Watsonx Integration Brings Enterprise AI Agent Orchestration
By
–
Excited to partner up with @IBM to have Watsonx integration in Flowise You can use IBM's foundational models, such as Granite, alongside other open-source models like Llama and Mistral. This marks a major step toward bringing AI Agent orchestration to enterprise settings.
-
Torch.compile() Outperforms Triton for Kernel Optimization
By
–
It's interesting that this "kernel" is designed for torch.compile(), so it's Python code but turns out faster! Conversely, some of the other Liger kernels are Triton and I measured them as slower than torch.compile'd versions.
-
Signal Releases New Feature for Enhanced User Experience
By
–
New, extremely welcome Signal feature just dropped! https://t.co/K83sV6ZPF7
— Meredith Whittaker (@mer__edith) 11 décembre 2024New, extremely welcome Signal feature just dropped!
-
BigData on Kubernetes: Building Efficient Scalable Data Solutions
By
–
#BigData on Kubernetes — A practical guide to building efficient and scalable data solutions: http://
amzn.to/3Xbir8T v/ @PacktPublishing —
#CDO #Analytics #DataScience #AI #Cloud #ML #DataEngineering
—
Key Features:
Leverage Kubernetes in a cloud environment to integrate -
Replit Agent Launches with Unlimited Usage and Visual Outputs
By
–
Announcing updates to the Replit Agent.
— Replit ⠕ (@Replit) 10 décembre 2024
The early access period is over. Today, we are launching:
1. Assistant
2. Unlimited usage
3. Attachments + URLs
4. Visual outputs
Replit Agent is the best tool for ANYONE — technical and non-technical alike — to build ideas 0 to 1. https://t.co/7azTO8jsiEAnnouncing updates to the Replit Agent. The early access period is over. Today, we are launching:
1. Assistant
2. Unlimited usage
3. Attachments + URLs
4. Visual outputs Replit Agent is the best tool for ANYONE — technical and non-technical alike — to build ideas 0 to 1.