I have a python code which uses keras and tensorflow backend. My system doesn't support training this model due to low memory space. I want to take use of Amazon sagemaker.
However all the tutorials I find are about deploying your model in docker containers. My model isn't trained and I want to train it on Amazon Sagemaker.
Is there a way to do this?
EDIT : Also can I make a script of my python code and run on it on AWS sagemaker?
SageMaker provides the capability for users to bring in their custom training scripts and train their algorithms using the script it on SageMaker using one of the pre-built containers for frameworks like TensorFlow, MXNet, PyTorch.
Please take a look at https://github.com/aws/amazon-sagemaker-examples/blob/master/frameworks/tensorflow/get_started_mnist_train.ipynb
It walks through how you can bring in your training script using TensorFlow and train it using SageMaker.
There are several other examples in the repository which will help you answer other questions you might have as you progress on with your SageMaker journey.
Related
Question is really clear. Nowadays im learning AWS world, and this question is eating my head up. What is the difference of import xgboost and import sagemaker.xgboost.
On SageMaker i can work with normal XGBoost library, and i know i can select different EC2 types with sagemaker.xgboost. But except this, what is the difference?
Are there any big difference?
Using model training as an example task: sagemaker.xgboost provides the ability to create Amazon SageMaker training jobs (and related AWS resources) in an environment that has the XGBoost library installed. So import xgboost gives you the modules for writing a training script that actually trains a model whereas import sagemaker.xgboost gives you modules for performing the training task on SageMaker.
The same applies for other tasks (e.g. predictions).
SageMaker XGBoost documentation: https://sagemaker.readthedocs.io/en/stable/frameworks/xgboost/using_xgboost.html#use-the-open-source-xgboost-algorithm
I'm trying to avoid migrating an existing model training process to SageMaker and avoid creating a custom Docker container to host our trained model.
My hope was to inject our existing, trained model into the pre-built scikit learn container that AWS provides via the sagemaker-python-sdk. All of the examples that I have found require training the model first which creates the model/model configuration in SageMaker. This is then deployed with the deploy method.
Is it possible to provide a trained model to the deploy method and have it hosted in the pre-built scikit learn container that AWS provides?
For reference, the examples I've seen follow this order of operations:
Creating an instance of sagemaker.sklearn.estimator.SKLearn and providing a training script
Call the fit method on it
This creates the model/model configuration in SageMaker
Call the deploy method on the SKLearn instance which automagically takes the model created in step 2/3 and deploys it in the pre-build scikit learn container as an HTTPS endpoint.
Yes, you can import existing models to SageMaker.
For scikit-learn, you would use the SKLearnModel() object to load to model from S3 and create it in SageMaker. Then, you could deploy it as usual.
https://sagemaker.readthedocs.io/en/latest/sagemaker.sklearn.html
Here's a full example based on MXNet that will point you in the right direction:
https://github.com/awslabs/amazon-sagemaker-examples/blob/master/sagemaker-python-sdk/mxnet_onnx_superresolution/mxnet_onnx.ipynb
Struggled with the same use case for a couple days.
We used sagemaker.model.Model class and sagemaker.pipeline.PipelineModel
Outlined our solution here.
How to handle custom transformation/ inference and requirements in sagemaker endpoints
I have already developed a scikit learn based machine learning model and have it in a pickle file. I am trying to deploy it only for inferencing and found sagemaker on aws. I do not see scikit learn based libraries on their available libraries and I also do not want to train the model all over again. Is it possible to only deploy the model that is already trained and present in AWS S3 on sagemaker?
You need to containerize it before deploy to SageMaker.
This might be a good start: https://aws.amazon.com/blogs/machine-learning/train-and-host-scikit-learn-models-in-amazon-sagemaker-by-building-a-scikit-docker-container/
I have a model built using python and tensorflow.
The model is trained and works well. I don't understand how I can deploy it? I mean how can I call this model in order to obtain a score on actual data?
I cannot use Watson ML deploy because of TensorFlow.
DSX supports training TensorFlow (without GPUs). I hear DSX will support training TensorFlow with GPUs and then deploying into Watch Machine Learning (WML) in early 2018.
For other models that you've built in DSX using SparkML, ScikitLearn, XGBoost and SPSS, go here for details on how deploy using WML:
Scala Jupyter Notebook end-to-end tutorial-Train and deploy a SparkML model
Python Jupyter Notebook end-to-end tutorial-Train and deploy a SparkML model
Python Jupyter Notebook: Recognition of hand-written digits--Train and deploy Scikit Learn model
From google tutorial we know how to train a model in TensorFlow. But what is the best way to save a trained model, then serve the prediction using a basic minimal python api in production server.
My question is basically for TensorFlow best practices to save the model and serve prediction on live server without compromising speed and memory issue. Since the API server will be running on the background for forever.
A small snippet of python code will be appreciated.
TensorFlow Serving is a high performance, open source serving system for machine learning models, designed for production environments and optimized for TensorFlow. The initial release contains C++ server and Python client examples based on gRPC. The basic architecture is shown in the diagram below.
To get started quickly, check out the tutorial.