sarvamai/amazon-sagemaker-examples

sarvamai/amazon-sagemaker-examples

Description: Sample notebooks and test data for models hosted on Amazon AWS Sagemaker

Language: Jupyter Notebook

Stars: 0

Forks: 0

Open issues: 0

Created: 2026-07-22T07:14:14Z

Pushed: 2026-07-22T15:21:31Z

Default branch: main

Fork: no

Archived: no

README:

Sarvam AI — Amazon SageMaker Examples

Sample notebooks and reference data for using Sarvam AI model packages from the AWS Marketplace with Amazon SageMaker.

Each notebook walks through the full lifecycle of a listing: subscribing to the model package, deploying a real-time endpoint, invoking it with the product's input/output schema, running batch transform jobs, and cleaning up.

Repository layout

Examples are grouped by product type, then by model (one folder per marketplace listing). Models of the same product type can have different input/output schemas, so every model folder is fully self-contained — its notebook, sample inputs, and expected outputs live together:

/ e.g. speech-to-text, vision, text-to-speech, llm, translate
└── / e.g. saaras-v3.1, ocr3b
├── -Model.ipynb sample notebook for the listing
├── README.md model overview + I/O schema summary
└── data/
├── input/ sample payloads used by the notebook
└── output/ expected responses for the sample inputs

Products whose payload format differs per inference mode (e.g. speech-to-text) further split data/input and data/output into real-time/, streaming/, and batch/ subfolders.

Available examples

| Product type | Model | Notebook | |---|---|---| | Speech-to-Text | Saaras v3.1 | [speech-to-text/saaras-v3.1](speech-to-text/saaras-v3.1/saaras-v3.1-speech-to-text-Model.ipynb) | | Vision (OCR) | OCR 3B | [vision/ocr3b](vision/ocr3b/ocr3b-vision-Model.ipynb) |

More product types (text-to-speech, LLM, translate) will be added as their marketplace listings launch.

Prerequisites

• An AWS account with an active subscription to the relevant marketplace listing (each notebook links to its listing).
• Amazon SageMaker Studio or a SageMaker Notebook Instance (recommended), with an execution role that has AmazonSageMakerFullAccess.
• Service quota for the GPU instance types named in each notebook.

Support

For questions about a listing or the models, contact Sarvam AI through the support channel listed on the marketplace product page.