GitLab

Today, we are thrilled to announce that DeepSeek R1 distilled Llama and Qwen models are available through Amazon and Amazon SageMaker JumpStart. With this launch, you can now deploy DeepSeek AI's first-generation frontier design, DeepSeek-R1, along with the distilled variations varying from 1.5 to 70 billion criteria to build, experiment, and responsibly scale your generative AI ideas on AWS.

In this post, we show how to get started with DeepSeek-R1 on Amazon Bedrock Marketplace and SageMaker JumpStart. You can follow comparable steps to release the distilled versions of the models also.

Overview of DeepSeek-R1

DeepSeek-R1 is a big language design (LLM) established by DeepSeek AI that uses support finding out to enhance reasoning abilities through a multi-stage training process from a DeepSeek-V3-Base foundation. A key identifying feature is its reinforcement learning (RL) step, which was utilized to refine the design's actions beyond the basic pre-training and tweak procedure. By incorporating RL, DeepSeek-R1 can adjust more efficiently to user feedback and goals, ultimately improving both significance and clearness. In addition, DeepSeek-R1 employs a chain-of-thought (CoT) approach, meaning it's geared up to break down intricate questions and reason through them in a detailed manner. This assisted reasoning process enables the design to produce more accurate, transparent, and detailed responses. This model combines RL-based fine-tuning with CoT capabilities, aiming to generate structured reactions while concentrating on interpretability and user interaction. With its extensive abilities DeepSeek-R1 has actually recorded the industry's attention as a versatile text-generation design that can be incorporated into numerous workflows such as representatives, logical thinking and information analysis tasks.

DeepSeek-R1 utilizes a Mixture of Experts (MoE) architecture and is 671 billion criteria in size. The MoE architecture allows activation of 37 billion specifications, enabling efficient inference by routing questions to the most pertinent expert "clusters." This technique permits the model to focus on various problem domains while maintaining total efficiency. DeepSeek-R1 needs at least 800 GB of HBM memory in FP8 format for inference. In this post, we will use an ml.p5e.48 xlarge instance to deploy the model. ml.p5e.48 xlarge comes with 8 Nvidia H200 GPUs offering 1128 GB of GPU memory.

DeepSeek-R1 distilled models bring the reasoning capabilities of the main R1 model to more efficient architectures based upon popular open designs like Qwen (1.5 B, 7B, 14B, and 32B) and Llama (8B and 70B). Distillation refers to a process of training smaller, more efficient models to simulate the habits and thinking patterns of the bigger DeepSeek-R1 design, utilizing it as a teacher design.

You can deploy DeepSeek-R1 model either through SageMaker JumpStart or Bedrock Marketplace. Because DeepSeek-R1 is an emerging design, we advise releasing this design with guardrails in place. In this blog, we will use Amazon Bedrock Guardrails to present safeguards, avoid damaging material, and examine designs against essential safety criteria. At the time of writing this blog, for DeepSeek-R1 implementations on SageMaker JumpStart and Bedrock Marketplace, Bedrock Guardrails supports only the ApplyGuardrail API. You can create numerous guardrails tailored to various usage cases and use them to the DeepSeek-R1 design, improving user experiences and standardizing safety controls across your generative AI applications.

Prerequisites

To deploy the DeepSeek-R1 design, you require access to an ml.p5e instance. To check if you have quotas for P5e, open the Service Quotas console and under AWS Services, choose Amazon SageMaker, and verify you're utilizing ml.p5e.48 xlarge for endpoint usage. Make certain that you have at least one ml.P5e.48 xlarge circumstances in the AWS Region you are deploying. To ask for a limitation increase, develop a limitation increase demand and reach out to your account team.

Because you will be releasing this model with Amazon Bedrock Guardrails, make certain you have the proper AWS Identity and Gain Access To Management (IAM) consents to use Amazon Bedrock Guardrails. For directions, see Establish permissions to use guardrails for material filtering.

Implementing guardrails with the ApplyGuardrail API

Amazon Bedrock Guardrails enables you to present safeguards, avoid harmful material, and evaluate designs against essential safety criteria. You can implement security steps for the DeepSeek-R1 design using the Amazon Bedrock ApplyGuardrail API. This enables you to apply guardrails to evaluate user inputs and design reactions deployed on Amazon Bedrock Marketplace and SageMaker JumpStart. You can produce a guardrail using the Amazon Bedrock console or the API. For the example code to produce the guardrail, see the GitHub repo.

The basic circulation includes the following actions: First, the system gets an input for the design. This input is then processed through the ApplyGuardrail API. If the input passes the guardrail check, it's sent out to the design for inference. After receiving the design's output, another guardrail check is applied. If the output passes this last check, it's returned as the result. However, if either the input or output is intervened by the guardrail, a message is returned showing the nature of the intervention and whether it happened at the input or output stage. The examples showcased in the following areas show reasoning using this API.

Deploy DeepSeek-R1 in Amazon Bedrock Marketplace

Amazon Bedrock Marketplace gives you access to over 100 popular, emerging, and specialized structure models (FMs) through Amazon Bedrock. To gain access to DeepSeek-R1 in Amazon Bedrock, complete the following steps:

1. On the Amazon Bedrock console, choose Model catalog under Foundation designs in the navigation pane. At the time of writing this post, you can utilize the InvokeModel API to conjure up the model. It doesn't support Converse APIs and other Amazon Bedrock tooling. 2. Filter for DeepSeek as a provider and pick the DeepSeek-R1 model.

The model detail page supplies essential details about the design's capabilities, rates structure, and implementation guidelines. You can find detailed use directions, consisting of sample API calls and code bits for combination. The model supports different text generation jobs, consisting of material creation, code generation, and question answering, using its support learning optimization and CoT thinking capabilities. The page likewise consists of implementation alternatives and licensing details to help you get going with DeepSeek-R1 in your applications. 3. To begin utilizing DeepSeek-R1, choose Deploy.

You will be prompted to configure the release details for DeepSeek-R1. The design ID will be pre-populated. 4. For Endpoint name, go into an endpoint name (in between 1-50 alphanumeric characters). 5. For Number of circumstances, go into a variety of instances (in between 1-100). 6. For example type, select your instance type. For optimum performance with DeepSeek-R1, a GPU-based circumstances type like ml.p5e.48 xlarge is recommended. Optionally, you can set up innovative security and facilities settings, consisting of virtual private cloud (VPC) networking, pediascape.science service role authorizations, and encryption settings. For a lot of utilize cases, the default settings will work well. However, for production deployments, you might want to examine these settings to line up with your organization's security and compliance requirements. 7. Choose Deploy to begin utilizing the design.

When the release is complete, you can test DeepSeek-R1's abilities straight in the Amazon Bedrock playground. 8. Choose Open in playground to access an interactive user interface where you can explore various prompts and change design criteria like temperature level and optimum length. When utilizing R1 with Bedrock's InvokeModel and Playground Console, utilize DeepSeek's chat template for optimum outcomes. For instance, content for reasoning.

This is an outstanding method to explore the design's reasoning and text generation abilities before integrating it into your applications. The play area provides instant feedback, assisting you understand how the design reacts to numerous inputs and letting you tweak your triggers for ideal outcomes.

You can rapidly evaluate the design in the play area through the UI. However, to invoke the deployed design programmatically with any Amazon Bedrock APIs, you need to get the endpoint ARN.

Run reasoning using guardrails with the released DeepSeek-R1 endpoint

The following code example demonstrates how to perform reasoning using a released DeepSeek-R1 design through Amazon Bedrock utilizing the invoke_model and ApplyGuardrail API. You can create a guardrail using the Amazon Bedrock console or the API. For the example code to develop the guardrail, see the GitHub repo. After you have developed the guardrail, utilize the following code to implement guardrails. The script initializes the bedrock_runtime customer, sets up inference criteria, and sends a demand to produce text based upon a user timely.

Deploy DeepSeek-R1 with SageMaker JumpStart

SageMaker JumpStart is an artificial intelligence (ML) hub with FMs, built-in algorithms, and prebuilt ML solutions that you can release with just a few clicks. With SageMaker JumpStart, you can tailor pre-trained models to your use case, with your data, and deploy them into production using either the UI or SDK.

Deploying DeepSeek-R1 design through SageMaker JumpStart offers 2 practical methods: using the instinctive SageMaker JumpStart UI or executing programmatically through the SageMaker Python SDK. Let's check out both approaches to assist you select the approach that best suits your needs.

Deploy DeepSeek-R1 through SageMaker JumpStart UI

Complete the following actions to release DeepSeek-R1 utilizing SageMaker JumpStart:

1. On the SageMaker console, choose Studio in the navigation pane. 2. First-time users will be prompted to develop a domain. 3. On the SageMaker Studio console, choose JumpStart in the navigation pane.

The model web browser shows available models, with details like the provider name and design abilities.

4. Search for DeepSeek-R1 to see the DeepSeek-R1 model card. Each design card shows key details, including:

- Model name

• Provider name
• Task classification (for example, Text Generation). Bedrock Ready badge (if applicable), showing that this model can be registered with Amazon Bedrock, enabling you to use Amazon Bedrock APIs to invoke the design
  
  5. Choose the design card to view the design details page.
  
  The model details page consists of the following details:
  
  - The design name and supplier details. Deploy button to release the design. About and Notebooks tabs with detailed details
  
  The About tab includes crucial details, such as:
  
  - Model description.
• License details.
• Technical specs.
• Usage guidelines
  
  Before you deploy the model, it's advised to review the design details and license terms to validate compatibility with your usage case.
  
  6. Choose Deploy to continue with deployment.
  
  7. For Endpoint name, utilize the instantly generated name or develop a custom-made one.

8. For Instance type ¸ pick a circumstances type (default: ml.p5e.48 xlarge).
9. For Initial circumstances count, go into the variety of instances (default: 1). Selecting proper circumstances types and counts is important for expense and efficiency optimization. Monitor your implementation to adjust these settings as needed.Under Inference type, Real-time inference is chosen by default. This is enhanced for sustained traffic and low latency.
10. Review all setups for precision. For this design, we highly suggest adhering to SageMaker JumpStart default settings and making certain that network seclusion remains in location.
11. Choose Deploy to release the design.
    
    The release process can take numerous minutes to finish.
    
    When release is total, your endpoint status will alter to InService. At this moment, the model is prepared to accept reasoning demands through the endpoint. You can keep track of the implementation progress on the SageMaker console Endpoints page, which will show relevant metrics and status details. When the implementation is complete, you can invoke the design utilizing a SageMaker runtime client and incorporate it with your applications.
    
    Deploy DeepSeek-R1 utilizing the SageMaker Python SDK
    
    To begin with DeepSeek-R1 using the SageMaker Python SDK, you will require to install the SageMaker Python SDK and make certain you have the required AWS authorizations and environment setup. The following is a detailed code example that shows how to deploy and utilize DeepSeek-R1 for reasoning programmatically. The code for deploying the design is provided in the Github here. You can clone the notebook and run from SageMaker Studio.
    
    You can run additional requests against the predictor:
    
    Implement guardrails and run inference with your SageMaker JumpStart predictor
    
    Similar to Amazon Bedrock, you can likewise utilize the ApplyGuardrail API with your SageMaker JumpStart predictor. You can create a guardrail using the Amazon Bedrock console or the API, and implement it as revealed in the following code:
    
    Clean up
    
    To prevent undesirable charges, complete the actions in this area to tidy up your resources.
    
    Delete the Amazon Bedrock Marketplace deployment
    
    If you released the design utilizing Amazon Bedrock Marketplace, total the following steps:
    
    1. On the Amazon Bedrock console, under Foundation models in the navigation pane, choose Marketplace deployments.
12. In the Managed deployments section, find the endpoint you want to delete.
13. Select the endpoint, and on the Actions menu, pick Delete.
14. Verify the endpoint details to make certain you're deleting the proper release: 1. Endpoint name.
15. Model name.
16. Endpoint status
    
    Delete the SageMaker JumpStart predictor
    
    The SageMaker JumpStart model you released will sustain expenses if you leave it running. Use the following code to delete the endpoint if you wish to stop sustaining charges. For more details, see Delete Endpoints and Resources.
    
    Conclusion
    
    In this post, we checked out how you can access and release the DeepSeek-R1 model using Bedrock Marketplace and SageMaker JumpStart. Visit SageMaker JumpStart in SageMaker Studio or Amazon Bedrock Marketplace now to get started. For more details, describe Use Amazon Bedrock tooling with Amazon SageMaker JumpStart designs, SageMaker JumpStart pretrained designs, Amazon SageMaker JumpStart Foundation Models, Amazon Bedrock Marketplace, and Starting with Amazon SageMaker JumpStart.
    
    About the Authors
    
    Vivek Gangasani is a Lead Specialist Solutions Architect for Inference at AWS. He helps emerging generative AI business develop innovative solutions using AWS services and accelerated calculate. Currently, he is focused on developing techniques for fine-tuning and optimizing the inference performance of large language models. In his leisure time, Vivek delights in treking, viewing motion pictures, and trying different cuisines.
    
    Niithiyn Vijeaswaran is a Generative AI Specialist Solutions Architect with the Third-Party Model Science team at AWS. His location of focus is AWS AI accelerators (AWS Neuron). He holds a Bachelor's degree in Computer technology and Bioinformatics.
    
    Jonathan Evans is a Professional Solutions Architect working on generative AI with the Third-Party Model Science team at AWS.
    
    Banu Nagasundaram leads item, engineering, and strategic partnerships for Amazon SageMaker JumpStart, SageMaker's artificial intelligence and generative AI center. She is passionate about constructing options that help clients accelerate their AI journey and unlock business worth.