Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1,309 changes: 1,309 additions & 0 deletions autopilot/autopilot_ts_data_merge.ipynb

Large diffs are not rendered by default.

Original file line number Diff line number Diff line change
@@ -0,0 +1,41 @@
# Amazon SageMaker Faster Autoscaling

To demonstrate newer, faster SageMaker autoscaling features, We deploy Meta's **Llama3-8B-Instruct** model to an Amazon SageMaker real-time endpoint using Text Generation Inference (TGI) Deep Learning Container (DLC).

To trigger autoscaling, we need to generate traffic to the endpoint.
We use [LLMPerf](https://github.com/philschmid/llmperf) to generate sample traffic to the endpoint.

## Prerequisites

Before using this notebook please ensure you have access to an active access token from HuggingFace and have accepted the license agreement from Meta.

- Step 1: Create user access token in HuggingFace (HF). Refer [here](https://huggingface.co/docs/hub/security-tokens) on how to create HF tokens.
- Step 2: Login to HuggingFace and navigate to [Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct/tree/main) home page.
- Step 3: Accept META LLAMA 3 COMMUNITY LICENSE AGREEMENT by following the instructions [here](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct/tree/main).
- Step 4: Wait for the approval email from META (Approval may take any where b/w 1-3 hrs).

---

>NOTE: LLMPerf spins up a ray cluster to generate traffic to Amazon SageMaker endpoint.\
>When running this on Amazon SageMaker Notebook Instance, ensure you use at least **m5.2xlarge** or a larger instance type.

## Autoscaling on real-time endpoints

### Amazon SageMaker real-time endpoints

- For Application Autoscaling example on Amazon SageMaker real-time endpoints refer to [FasterAutoscaling-SME-Llama3-8B-AppAutoScaling.ipynb](./realtime-endpoints/FasterAutoscaling-SME-Llama3-8B-AppAutoScaling.ipynb) notebook.

- For StepScaling example on Amazon SageMaker real-time endpoints refer to [FasterAutoscaling-SME-Llama3-8B-StepScaling.ipynb](./realtime-endpoints/FasterAutoscaling-SME-Llama3-8B-StepScaling.ipynb) notebook.

### Amazon SageMaker Inference Components

- For autoscaling example using Amazon SageMaker Inference components, refer to [inference-component-llama3-autoscaling.ipynb](./realtime-endpoints/FasterAutoscaling-IC-Llama3-8B-AppAutoScaling.ipynb) notebook.

---

## References

- [LLMPerf](https://github.com/philschmid/llmperf)
- [Llama3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct)
- [Create HF Access Token](https://huggingface.co/docs/hub/security-tokens)
- [Amazon SageMaker Inference Components - blog post](https://aws.amazon.com/blogs/machine-learning/reduce-model-deployment-costs-by-50-on-average-using-sagemakers-latest-features/)
Loading