FALCON 3 models are now available at Amazon Sagemaker Jumpstart

Today, we are very happy to announce that Family 3 Family of Models from TII is available at Amazon Sagemaker JumpStart. In this case, we check how we can use this model well in Amazon Sagemaker AI.

View all Falcon 3 Family of Models

FALCON 3 FAMILY, developed by Technology Innovation Institute (TII) in Abu Dhabi, representing important development in open source models. This group includes five sessions range from 1 to 10 parameters, focusing on improving science, mathematical, and coding skills. The family contains FALCON3 5

These models show new things such as appropriate training strategies before training, measure advanced thinking, and information entries to function better in smaller models. Significantly, FALCON3-10B-Based model achieves the operation of the models of the models under 13 billion billions of zero-shots with a few shots of shooting. Falcon 3 family also includes various versions of teaching as teaching and supports different measurement formats, making them converted to many applications.

Currently, the Sagemaker JumpStart offers the Falcon3-3b versions, FALCON3-7B, and Falcon3-10B, and their corresponding education.

Get started with Sagemaker JumpStart

Sagemaker JumpStart is a Machine Learning Organization (ML) that can help speed up your ML trip. With Sagemaker JumpStart, you can view, and compare, and select Pre-Processed Basel Processed (FMS), including Falcon 3 models. These types are completely customized for your data.

Moving Falcon 3 model with Sagemaker JumpStart offers two simple ways: Using Sagemaker Jumpstart Sagemaker Ui or to use the Sagemaker Python SDK. Let us consider both ways to help you choose a method that suits your needs.

Send Falcon 3 using Sagemaker JumpStart Ui

Complete the following steps to install FALCON 3 through Jumpstart UI:

1. To access the Sagemaker JumpStart, use one of the following methods:
   1. In the Amazon Sagemaker Unified Studio, at Live Change the menu Jumpstart models behind Development Model.
   2. Alternatively, Amazon Sagemaker Studio, Select Bend to the wavering pane.
      

2. Search FALCON3-10B-BASE in the model browser.
   
3. Select the model and select Deploy.
   
4. A Members Type Typeor use a default example or select a different example.
5. Designate Deploy.
   After a while, the ENDPOINT state will show as Setting a Center Insice And you will be able to move the influence you should.
   

Send Failcon 3 through Sagemaker Python SDK

In groups monitor exchange of shipment or encounter with existing mlops pipes, you can use Sagemaker Python SDK:

from sagemaker.serve.builder.model_builder import ModelBuilder
from sagemaker.serve.builder.schema_builder import SchemaBuilder
from sagemaker.jumpstart.model import ModelAccessConfig
from sagemaker.session import Session
import logging

sagemaker_session = Session()

artifacts_bucket_name = sagemaker_session.default_bucket()
execution_role_arn = sagemaker_session.get_caller_identity_arn()


js_model_id = "huggingface-llm-falcon-3-10B-base"

gpu_instance_type = "ml.g5.12xlarge"  

response = "Hello, I'm a language model, and I'm here to help you with your English."

sample_input = {
    "inputs": "Hello, I'm a language model,",
    "parameters": {"max_new_tokens": 128, "top_p": 0.9, "temperature": 0.6},
}

sample_output = [{"generated_text": response}]

schema_builder = SchemaBuilder(sample_input, sample_output)

model_builder = ModelBuilder(
    model=js_model_id,
    schema_builder=schema_builder,
    sagemaker_session=sagemaker_session,
    role_arn=execution_role_arn,
    log_level=logging.ERROR
)

model= model_builder.build()

predictor = model.deploy(model_access_configs={js_model_id:ModelAccessConfig(accept_eula=True)}, accept_eula=True)

Run the forecast for a predictable:

predictor.predict(sample_input)

If you want to set up the ability to scope down to zero after shipping, refer to saving costs at the new rate on the Sagemaker area.

Clean

Cleaning model and Endpoint, use the following code:

predictor.delete_model()
predictor.delete_endpoint()

Store

In this sentence, we checked that Sagemaker JumpStart provides Scientific Data and ML Engineers to find, access, and use a broad range of pre-acquisition FALCON 3 FALCON 3 FALCON OF MODELS. Visit Sagemaker JumpStart Sagemaker Studio now to get started. For more information, see Sagimaker JumpStart models, Amazon Sagemaker JumpStart Foundal Model, and start with Amazon Sagemaker JumpStart.

About the authors

Niithyn Vijeaweran Is the construction of the solutions of AI generating and a third Model Model team in AWS. His focus facility is Aisceter for AISCEATER AWS AWS. Hold a bachelor degree in computer science and biooinformatics.

Marc Karp Is ML builder and amazon sagemaker service group. You focus on helping customers design, put, and treat ML loads on a scale. In his spare time, she enjoys walking and exploring new places.

Raghu Ramesha Is the ML solutions builder and Amazon Sagemaker service group. You focus on helping customers to build, put, and move loads of ML production work in Sagomaker on a scale. You especially work in a machine learning, AI, and the computer vision Domain, and holds a Master degree in computer science from UT Dallas. In his free time, she enjoys walking and pictures.

Baby Nagasundaram It leads the product, engineering, and the Sagemaker JumpStart's strategies, study and Genai Hub machine. You are passionate about building solutions that help customers accelerate their AI and open the business value.