We're thrilled to unveil a significant addition to the Azure AI model catalog at Microsoft Build 2024—the integration of Nixtla's TimeGEN-1time-series forecasting model. This model will be offered as a Model as a Service (MaaS), and it is set to how businesses forecast future events across various industries. Microsoft Azure is the first cloud provider to offer a foundation model for this time-series model.
Azure AI Model Catalog with Nixtla TimeGEN-1's announcement
Nixtla is renowned for its groundbreaking time-series forecasting model, TimeGPT. Nixtla is pioneering in time-series models, similar to what OpenAI and others have achieved for language models – making them accessible to anyone. TimeGPT has been optimized for the Azure architecture and is now presented as TimeGEN-1 in the Azure AI model catalog.
TimeGEN-1: Advanced Time Series Forecasting Model
Time-series forecasting is a critical domain that involves predicting future values based on previously observed data points. It’s widely used across various industries, from finance and retail to healthcare and climate science. However, this field comes with its own set of challenges such as seasonality, outliers, trends, non-stationarity, and missing data.
Foundation models, such as deep learning architectures, have begun to address these challenges by leveraging large amounts of data and advanced computational techniques. They are pre-trained on vast datasets, allowing them to recognize complex patterns and dependencies that traditional models might miss.
Nixtla's TimeGEN-1 on Azure AI model catalog
Nixtla’s TimeGEN-1 model is a state-of-the-art generative pre-trained transformer foundation model designed specifically for time-series forecasting. It’s a powerful tool that can produce accurate forecasts from historical data without the need for retraining for each specific task - which means users can get started out of the box without the need for machine learning engineers. TimeGEN-1 treats time-series forecasting in the same way how natural language processing (NLP) models handle text—by "reading" a sequence of data points (or "tokens") and predicting future values based on learned patterns. TimeGEN-1 also stands out for its ability to fine-tune with your own data, offering anomaly detection and low latency in its operations. This model can democratize access to advanced predictive insights, assisting both individuals and organizations to navigate uncertainty and make data-driven decisions with ease. Whether you’re forecasting market trends or predicting product demand, TimeGEN-1 can simplify the process, making cutting-edge time series analysis accessible to all.
Integrating TimeGEN-1 with Azure AI not only extends its accessibility but also can enrich the forecasting experience with enhanced features and tools. Azure AI Studio and Azure Machine Learning facilitate easy model management and deployment, enabling users to swiftly move from data ingestion to insight generation.
"We are immensely proud to bring TimeGEN-1 into the Azure AI ecosystem, partnering with Microsoft to redefine how businesses approach time-series forecasting,” said Max Mergenthaler Canseco, CEO and Co-founder at Nixtla. “This collaboration is a significant milestone for Nixtla, as it combines our advanced forecasting technology with the robust and scalable Azure AI platform. Our goal is to democratize access to powerful AI tools, and together with Azure, we are turning this vision into reality by enabling organizations to deploy high-accuracy, cost-effective forecasting solutions at scale."
The introduction of TimeGEN-1 into Azure AI marks a significant enhancement in how businesses can harness advanced AI for time series forecasting. This integration is built on several key pillars that ensure both robust functionality and adherence to best practices in AI deployment:
- Enhanced Security and Compliance: Azure AI prioritizes the security and privacy of user data, incorporating Microsoft's comprehensive security protocols. TimeGEN-1 operates within this secure framework, ensuring that all data handled by the model is protected against unauthorized access and threats. This setup helps enterprises maintain high standards of data privacy and operational security, fostering trust and compliance. You can learn more here.
- Simplified Deployment and Inference: TimeGEN-1 leverages Azure AI's robust infrastructure to offer a simplified deployment process. Users can deploy the Model as a Service (MaaS), utilizing Azure’s managed services for scalable, pay-as-you-go inference without the complexities of managing the underlying hardware. This ease of deployment accelerates the operational rollout and reduces the technical barrier to advanced analytics.
- End-to-end Operationalization with MLOps: The integration of TimeGEN-1 with Azure Machine Learning pipelines includes comprehensive support for lifecycle management of the model through MLOps (Machine Learning Operations). This approach encompasses everything from deployment to monitoring and management, ensuring that the model and ML workflows remains efficient and effective throughout its operational life.
This integration represents a forward-thinking approach to enterprise AI deployment, empowering businesses to leverage cutting-edge technology while maintaining rigorous standards of security and compliance.
Customers are already using TimeGEN-1 on Azure!
TimeGEN-1 is already revolutionizing the way businesses across various industries handle their predictive analytics. From enhancing demand forecasting in retail and manufacturing to optimizing financial predictions in investment research, TimeGEN-1 on Azure empowers organizations to achieve unparalleled accuracy and efficiency. MindsDB, a leading AI Startup, leverages TimeGEN-1 to enable their customers to perform rapid and precise forecasting across diverse applications such as anomaly detection and large-scale predictions, drastically reducing complexity and time investment. Similarly, OpenBB Terminal Pro integrates TimeGEN-1 to allow financial analysts and quants to effortlessly generate forecasts from proprietary datasets, thus democratizing access to advanced forecasting technologies.
In the life sciences sector, RoadMap Technologies incorporates TimeGEN-1 within its TrailBlazer platform, providing users with robust and integrated forecasting solutions that quantify uncertainty and enhance decision-making. STIHL, a global leader in power equipment, utilizes TimeGEN-1 to optimize its inventory and production processes, achieving significant improvements in forecasting accuracy for its extensive product lineup. These diverse applications underscore the versatility and transformative potential of TimeGEN-1, making state-of-the-art forecasting accessible to companies of all sizes and across various sectors.
At Bridgestone we value customers, and to make sure that we provide our customers tires at the right time we need to optimize the upstream. To do so we are working on state-of-the-art forecasting models. In this regards we value our partnership with Nixtla and Microsoft.
- Onkar Ambekar, Director AI & Analytics (EMEA & Americas), Bridgestone.
Before TimeGEN-1, our team spent a lot of time creating and maintaining forecasting pipelines. Now, we do state-of-the-art forecasting in a few lines of code and in just a couple of seconds. TimeGEN-1 saved us major hours and headaches.
- Mark Angler, Business Intelligence Team Lead, STIHL
As the leading predictive analytics models in the market, TimeGEN-1 offers advanced capabilities that provide a variety of unique features, making it a powerful asset for managing complex forecasting scenarios. Integrating TimeGEN-1 with MindsDB creates an impactful combination for predictive insights directly within business databases, so organizations can react swiftly to a rapidly evolving global market.
- Jorge Torres, CEO, MindsDB
Get started with TimeGEN-1 on Azure AI
Here are the prerequisites:
- If you don’t have an Azure subscription, get one here: https://azure.microsoft.com/en-us/pricing/purchase-options/pay-as-you-go
- Create an Azure AI Studio hub and project. Supported regions are: East US 2, Sweden Central, North Central US, East US, West US, West US3, South Central US. Make sure you pick one these as the Azure region for the hub.
Next, you need to create a deployment to obtain the inference API and key:
- Open the TimeGEN-1 model card in the model catalog: https://aka.ms/aistudio/landing/nixtlatimegen1
- Click on Deploy and select the Pay-as-you-go option.
- Subscribe to the Marketplace offer and deploy. You can also review the API pricing at this step.
- You should land on the deployment page that shows you the API and key in less than a minute.
Follow this article to learn more about TimeGEN-1.
FAQ
- What does it cost to use the TimeGEN-1 model on Azure?
- You are billed based on the number of input and output tokens. You can review the pricing in the Marketplace offer details tab when deploying the model. You can also find the pricing on the Azure Marketplace.
- Do I need GPU capacity in my Azure subscription to use TimeGEN-1?
- No, you do not need GPU capacity. The TimeGEN-1 is offered as an API through Models as a Service.
- Is TimeGEN-1 available in Azure Machine Learning Studio?
- Yes, TimeGEN-1 is available on model catalog in both Azure AI Studio and Azure Machine Learning Studio.
- TimeGEN-1 is listed on the Azure Marketplace. Can I purchase and use TimeGEN-1 directly from Azure Marketplace?
-
- Azure Marketplace is our foundation for commercial transactions for models built on or built for Azure. The Azure Marketplace enables the purchasing and billing of TimeGEN-1. However, model discoverability occurs in both Azure Marketplace and the Azure AI Model Catalog. Meaning which, users can search and find Time-GEN1 in both the Azure Marketplace and Azure AI Model Catalog.
- If the user searches for TimeGEN-1 in Azure Marketplace, they're able to subscribe to the offer before being redirected to the Model Catalog in Azure AI Studio where they complete subscribing and can deploy the model.
- If the user searches for TimeGEN-1 in the Azure AI Model Catalog, they're able to subscribe and deploy the model from the Catalog without starting from the Azure Marketplace. The Azure Marketplace still tracks the underlying commerce flow.
-
- Given that TimeGEN-1 is billed through the Azure Marketplace, does it retire my Azure consumption commitment (aka MACC)?
- Yes, TimeGEN-1 is an “Azure benefit eligible” Marketplace offer, which indicates MACC eligibility. Learn more about MACC here: https://learn.microsoft.com/en-us/marketplace/azure-consumption-commitment-benefit
- Is my inference data shared with TimeGEN-1?
- No, Microsoft does not share the content of any inference request or response data with Nixtla. Your data, including the data generated through your organization’s use of Models as a Service on Azure – such as prompts and responses – are kept private and are not disclosed to third parties. Microsoft also doesn't use prompts and outputs to train nor improve any MaaS models. Additionally, all MaaS models are stateless, which means no prompts or outputs are stored in the models. You can learn more here.
- Are there rate limits for the TimeGEN-1model on Azure?
- Yes, there are rate limits for the TimeGEN-1model on Azure. Each deployment has a rate limit of 200,000 tokens per minute and 1,000 API requests per minute. Contact Azure customer support if you have additional questions.
- Is the TimeGEN-1 model region specific?
- TimeGEN-1 model API endpoints can be created in AI Studio projects to Azure Machine Learning workspaces in East US 2, Sweden Central, North Central US, East US, West US, West US3 or South Central US. Essentially, you can use the API from any Azure region once you create it in East US 2, Sweden Central, North Central US, East US, West US, West US3 or South Central US.
- Can I fine-tune the TimeGEN-1 model?
- TimeGEN-1 provides a finetuning parameter that can be used to finetune the model and generate forecasts in the same step. The finetuned model weights do not persist. The full finetuning experience is currently unavailable.
-
Can I use MaaS models in any Azure subscription types?
-
Customers can use MaaS models in all Azure subsection types with a valid payment method, except for the CSP (Cloud Solution Provider) program. Free or trial Azure subscriptions are not supported.
-