Sold by: FriendliAI
Deployed on AWS
Free Trial
Efficient, fast, and reliable generative AI inference solution of Llama 3.1 8B Instruct Int8 model
Overview
Imagine there is a powerful racecar (a generative AI model) that needs much maintenance and tuning (infrastructure and technical know-how). Friendli Container in SageMaker is like a rental service, taking care of the hassle so you can just drive! It provides a simple interface that connects you to Friendli Engine, a high-performance, cost-effective inference serving engine optimized for generative AI models.
This product is built with Llama. Llama 3.1 is licensed under the Llama 3.1 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved. The license is available at: https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/LICENSE
Highlights
- Access popular open-source models: Get started with pre-loaded models(Llama 3.1 8B Instruct). No need to worry about downloading or optimizing them.
- Build your own workflows: Integrate these models into your applications with just a few lines of code. Generate creative text formats, code, musical pieces, email, letters, etc. and create stunning images with ease.
- Focus on what matters: Forget about infrastructure setup and GPU optimization. Friendli Container handles the heavy lifting, freeing you to focus on your creative vision and application development.
Details
Sold by
Delivery method
Latest version
Deployed on AWS
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Trust Center
Access real-time vendor security and compliance information through their Trust Center powered by Drata or Vanta. Review certifications and security standards before purchase.
Financing for AWS Marketplace purchases
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Pricing
Free trial
Try this product free for 7 days according to the free trial terms set by the vendor.
Pricing is based on actual usage, with charges varying according to how much you consume. Subscriptions have no end date and may be canceled any time.
Additional AWS infrastructure costs may apply. Use the AWS Pricing Calculator to estimate your infrastructure costs.
Dimension | Description | Cost/host/hour |
|---|---|---|
ml.g5.xlarge Inference (Batch) Recommended | Model inference on the ml.g5.xlarge instance type, batch mode | $0.15 |
ml.g5.xlarge Inference (Real-Time) Recommended | Model inference on the ml.g5.xlarge instance type, real-time mode | $0.15 |
ml.g5.8xlarge Inference (Real-Time) | Model inference on the ml.g5.8xlarge instance type, real-time mode | $0.15 |
ml.g5.2xlarge Inference (Real-Time) | Model inference on the ml.g5.2xlarge instance type, real-time mode | $0.15 |
ml.g5.4xlarge Inference (Real-Time) | Model inference on the ml.g5.4xlarge instance type, real-time mode | $0.15 |
ml.g5.16xlarge Inference (Real-Time) | Model inference on the ml.g5.16xlarge instance type, real-time mode | $0.15 |
Dimensions summary
You pay by the hour for each host running model inference, billed only while the instance is active. Pricing is organized by AWS GPU instance size, from ml.g5.xlarge up through ml.g5.16xlarge. Choosing a larger instance size adds more compute capacity and changes the hourly rate. The ml.g5.xlarge instance offers two processing modes: batch mode, which groups requests together, and real-time mode, which responds to requests as they arrive. All other instance sizes run in real-time mode only. You scale cost by selecting an instance size and mode that match your workload.
Top-of-mind questions for buyers
One host hour covers one running AWS ml.g5 GPU instance serving the model for one hour. Instance sizes range from ml.g5.xlarge up to ml.g5.16xlarge, adding more GPU compute at each size. You pick the size that fits your inference load and pay per active host hour.
You pay only while the host runs. Charges accrue per active host hour, so a stopped instance does not incur software charges. Note that underlying AWS storage or reservation fees may still apply separately, but the software meters running time only.
Both meter per host hour. Batch mode groups incoming requests together before processing, suited to jobs that tolerate delay. Real-time mode responds to each request as it arrives, suited to interactive workloads. The ml.g5.xlarge offers both; all other sizes run real-time mode only.
friendli.ai+1
Helpful?
Vendor refund policy
We do not support any refunds currently.
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA) .
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Delivery details
Amazon SageMaker model
An Amazon SageMaker model package is a pre-trained machine learning model ready to use without additional training. Use the model package to create a model on Amazon SageMaker for real-time inference or batch processing. Amazon SageMaker is a fully managed platform for building, training, and deploying machine learning models at scale.
Deploy the model on Amazon SageMaker AI using the following options:
Version release notes
Initial version: Llama-3.1-8B-Instruct Int8
Additional details
Inputs
- Summary
The input request payloads that are compatible with OpenAI's Chat Completion endpoint.
- Limitations for input type
- For input and generated output tokens, a maximum context length is 128k
- Input MIME type
- text/csv, application/json, application/jsonlines
{
"messages": [
{
"role": "system",
"content": "You are a friendly AI assistant."
},
{
"role": "user",
"content": "Please explain Python language."
}
],
"temperature": 0.7
}
https://github.com/friendliai/examples/tree/main/aws/sagemaker/input.json
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
messages | A list of messages comprising the conversation so far.
Array[role, content, name]
role(string): The role of the messages author. Possible values: [system, user]
content(string): The content of message.
name(string): The name for the participant to distinguish between participants with the same role. | Type: FreeText
Limitations: Role must be the one of [system, user] | Yes |
frequency_penalty | Number between -2.0 and 2.0. Positive values penalizes tokens that have been sampled, taking into account their frequency in the preceding text. This penalization diminishes the model's tendency to reproduce identical lines verbatim. | Default value: null
Type: Continuous
Minimum: -2.0
Maximum: 2.0 | No |
presence_penalty | Number between -2.0 and 2.0. Positive values penalizes tokens that have been sampled at least once in the existing text. | Default value: null
Type: Continuous
Minimum: -2.0
Maximum: 2.0 | No |
repetition_penalty | Penalizes tokens that have already appeared in the generated result (plus the input tokens). should be greater than or equal to 1.0. 1.0 means no penalty. This is similar to Hugging Face transformer's repetition_penalty argument. | Default value: null
Type: Continuous
Minimum: 1.0 | No |
max_tokens | The maximum number of tokens to generate. The length of your input tokens plus max_tokens should not exceed the model's maximum length. This is similar to Hugging Face transformer's max_new_tokens argument. | Default value: null
Type: Integer | No |
n | The number of independently generated results for the prompt. Not supported when using beam search. Defaults to 1. This is similar to Hugging Face transformer's num_return_sequences argument. | Default value: 1
Type: Integer
Minimum: 1 | No |
stop | When one of the stop phrases appears in the generation result, the API will stop generation. The stop phrases are excluded from the result. | Default value: null
Type: FreeText
Limitations: string list | No |
temperature | Sampling temperature. Smaller temperature makes the generation result closer to greedy, argmax (i.e., top_k = 1) sampling. defaults to 1.0. This is similar to Hugging Face transformer's temperature argument. | Default value: 1.0
Type: Continuous | No |
top_p | Tokens comprising the top top_p probability mass are kept for sampling. Numbers between 0.0 (exclusive) and 1.0 (inclusive) are allowed. Defaults to 1.0. This is similar to Hugging Face transformer's top_p argument. | Default value: 1.0
Type: Continuous | No |
top_k | The number of highest probability tokens to keep for sampling. Numbers between 0 and the vocab size of the model (both inclusive) are allowed. The default value is 0, which means that the API does not apply top-k filtering. This is similar to Hugging Face transformer's top_k argument. | Default value: null
Type: Integer
Minimum: 1 | No |
Resources
Vendor resources
Support
Vendor support
FriendliAI has been developing key technologies to enable serving generative AI models. Friendli, our flagship generative AI serving engine is available to use.
If you have any problems with the form, please contact us here.
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Similar products
Llama-3-Varco-Offsetbias-8B is NC's generative judge model that performs pairwise preference evaluation task.
Available as a NIM, Llama 3.1 8B-Instruct is an 8B-parameter LLM pretrained, instruction tuned model optimized for multilingual dialogue.
Available as a NIM, Llama 3.1 70B-Instruct is a 70B-parameter LLM pretrained, instruction tuned model optimized for multilingual dialogue.
By Arcee
Llama-Spark is a conversational model developed by Arcee.ai. It excels at engaging in natural and informative conversations.
Llama-VARCO-8B-Instruct is a generative model built with Llama, specifically designed to excel in Korean through additional training
Customer reviews
No customer reviews yet
Be the first to review this product . We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.