Sold by: Karakuri, Inc.
Deployed on AWS
mixture of experts (MoE) architecture that supports multiple languages, primarily English and Japanese
Overview
KARAKURI Inc. has developed a large language model called "KARAKURI LM 8x7B Instruct v0.1" using a mixture of experts (MoE) architecture that supports multiple languages, primarily English and Japanese, and is licensed under Apache 2.0. The model is a finetuned version of tokyotech-llm/Swallow-MX-8x7b-NVE-v0.1 and can be used for various tasks, including generating responses to user prompts, internet search, and retrieving relevant information from a database. The model's performance is evaluated using nine attributes, including helpfulness, correctness, coherence, complexity, verbosity, quality, toxicity, humor, and creativity. The model was trained on approximately 1 billion tokens of fine-tuning data, and its known limitations include the possibility of attempting to call unprovided tools.
Highlights
- 1. **Support for Function calling and RAG**: KARAKURI LM is the first domestic LLM to support Function calling and RAG. This enables the use of external tools and databases to collect and analyze information, allowing for the execution of more advanced tasks.
- 2. **Learning optimized for business implementation**: KARAKURI LM prioritizes business implementation in its development. This results in a model with high practicality among domestic models, suitable for immediate use in Japanese business scenarios.
- 3. **Development at low cost**: KARAKURI LM utilizes AWS Trainium to reduce development costs by up to 50%. This enables even medium and small-sized enterprises and ventures to develop large-scale language models.
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
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
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.g5.48xlarge Inference (Batch) Recommended | Model inference on the ml.g5.48xlarge instance type, batch mode | $2.00 |
ml.g6.48xlarge Inference (Real-Time) Recommended | Model inference on the ml.g6.48xlarge instance type, real-time mode | $2.00 |
ml.p4de.24xlarge Inference (Real-Time) | Model inference on the ml.p4de.24xlarge instance type, real-time mode | $2.00 |
ml.p4d.24xlarge Inference (Real-Time) | Model inference on the ml.p4d.24xlarge instance type, real-time mode | $2.00 |
ml.g5.48xlarge Inference (Real-Time) | Model inference on the ml.g5.48xlarge instance type, real-time mode | $2.00 |
ml.p5.48xlarge Inference (Real-Time) | Model inference on the ml.p5.48xlarge instance type, real-time mode | $2.00 |
Dimensions summary
You pay by the hour for each instance you run (HostHrs), with no upfront commitment. Pricing depends on two choices. First, the inference mode: real-time serves live requests, while batch processes grouped jobs. Batch is offered only on the ml.g5.48xlarge instance. Second, the instance type, which spans five hardware options: ml.g5.48xlarge, ml.g6.48xlarge, ml.p4d.24xlarge, ml.p4de.24xlarge, and ml.p5.48xlarge. Each instance type carries its own hourly rate. You select the combination that fits your workload, and charges accrue only while the instance runs.
Top-of-mind questions for buyers
Billing meters each running instance by the hour (HostHrs). Charges accrue only while the instance is active. If you stop the instance, software charges stop. Real-time instances keep running to serve live requests, so they accrue hours continuously until you shut them down.
Real-time mode keeps an instance running to serve live requests, so you pay for every hour it stays up. Batch mode processes grouped jobs, so the instance runs only during the job. Batch is offered only on the ml.g5.48xlarge instance type here.
Each dimension maps to one AWS GPU instance size (g5, g6, p4d, p4de, or p5, 24xlarge or 48xlarge). You pay per hour for that hardware. The model itself can also run in other cloud, on-premises, or local environments, but those options are not billed through this listing.
ad.karakuri.ai
Helpful?
Vendor refund policy
Our refund policy strictly adheres to the conditions outlined in the End User License Agreement (EULA). Refunds will only be provided under the specific circumstances detailed in the EULA, such as when the product fails to function as described and corrective measures are not successfully taken within the designated timeframe.
Refunds will not be issued for any other circumstances beyond what is explicitly stated in the EULA.
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
This release updates the TGI version to v2.4.0.
Additional details
Inputs
- Summary
Please refer to the model card for details on prompt formatting and the API reference for API specifications.
- Input MIME type
- application/json
https://github.com/karakuri-ai/sagemaker-examples/blob/main/examples/karakuri-lm-8x7b-instruct-v0.1/sample-input.json
https://github.com/karakuri-ai/sagemaker-examples/blob/main/examples/karakuri-lm-8x7b-instruct-v0.1/sample-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 |
|---|---|---|---|
inputs | Text input for the model to respond to. | Type: FreeText | Yes |
Resources
Vendor resources
Support
Vendor support
For questions and comments about the model, please email karakuri-rd@karakuri.ai
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
By Arcee
Llama-Spark is a conversational model developed by Arcee.ai. It excels at engaging in natural and informative conversations.
Enterprise-grade 10B Japanese LLM, powered by AWS Trainium, delivering high accuracy, cost-efficient deployment, and seamless fine-tuning.
This is a repackaged open source software product wherein additional charges apply for technical support. Softwares Included: Red Hat Enterprise Linux 8 and Docker
This Rocky Linux 8 LVM image has charges associated with it for security hardening and optimization provided by ProComputers. It uses LVM (Logical Volume Manager), so storage can be resized and volumes expanded without downtime, giving developers and administrators flexibility and control in AWS environments. With Logical Volume Manager, you can grow volumes online, rearrange disk capacity, and manage storage layers without rebuilding instances. Built on Rocky Linux 8, the image offers RHEL 8 binary compatibility, SELinux in enforcing mode, ENA networking, and cloud-init for automated provisioning. Rocky 8 is a dependable base for web applications, databases, and CI/CD workloads, and ProComputers tests and maintains the image with regular updates for long-running production use on EC2.
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.