forked from intel-staging/oneAPI-samples
-
Notifications
You must be signed in to change notification settings - Fork 0
Samples for Intel oneAPI toolkits
bdattoma/oneAPI-samples
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"\n",
"# E2E samples to get started with Intel® oneAPI AI analytics toolkit\n",
"\n",
"This RHODS environment with Intel® oneAPI AI analytics toolkit aims to deliver high-performance DL workload on Intel® XPUs with Intel-optimized TensorFlow* and PyTorch*, Intel Model Zoo, and LPOT(Low precision optimization tool). The toolkit also includes drop-in acceleration for data preprocessing and machine learning workflows with compute-intensive Python* packages: Modin*, scikit-learn*, daal4py and XGBoost*.\n",
"\n",
"This Jupyter hub environment also features e2e samples to get started with understanding how Intel® oneAPI AI analytics toolkit delivers optimized solutions for DL and ML workflows. Below are the list of samples:\n",
"\n",
"1. **[E2E_use_case_with_Intel_optimized_Tensorflow_LPOT](./E2E_use_case_with_Intel_optimized_Tensorflow_LPOT)**: This sample utilizes Intel-optimized Tensorflow and LPOT( Low Precision Optimizations Tool) in the Intel® oneAPI API analytics toolkit offered in the RHODS environment. The sample will train MNIST with Intel-optimized Tensorflow on alexnet, followed by quantizing with LPOT to convert fp32 trained model to int8 low-precision model and perform optimized inference. This sample will also provide performance and accuracy comparisons on fp32 vs int8 inference highlighting the importance of using LPOT in Intel® oneAPI AI analytics toolkit to perform low-precision inference. Open the [lpot_sample_tensorflow.ipynb](./E2E_use_case_with_Intel_optimized_Tensorflow_LPOT/lpot_sample_tensorflow.ipynb) notebook and follow the instructions.\n",
"\n",
"2. **[E2E_use_case_with_Intel_Modin_Intel_optimized_scikit-learn](./E2E_use_case_with_Intel_Modin_Intel_optimized_scikit-learn)**: This sample will introduce users to trivial extensions of using Intel Modin and Intel® Extension for scikit-learn from the Intel® oneAPI API analytics toolkit offered in the RHODS environment. The sample trains US Census data with Intel® Extension for scikit-learn and utilizes Intel Modin on Pandas to perform optimized data preprocessing calls such as read_csv and other ETL operations. Open the [census_modin.ipynb](./E2E_use_case_with_Intel_Modin_Intel_optimized_scikit-learn/census_modin.ipynb) notebook and follow the instructions.\n",
"\n",
"3. **[E2E_use_case_with_Intel_optimized_XGBoost_daal4py](./E2E_use_case_with_Intel_optimized_XGBoost_daal4py)**: This sample utilizes the Intel performant XGBoost package (> 1.0 version) to train on higgs dataset, and daal4py package for additional acceleration to run predictions. Open the [IntelPython_XGBoost_daal4pyPrediction.ipynb](./E2E_use_case_with_Intel_optimized_XGBoost_daal4py/IntelPython_XGBoost_daal4pyPrediction.ipynb) notebook and follow the instructions.\n",
"\n",
"4. Intel Model Zoo is also shipped as part of the toolkit and can be found in the \"models\" folder. Go to models/quickstart on how to run various models offered as part of Intel Model Zoo \n",
"\n",
"For more samples, goto https://github.com/oneapi-src/oneAPI-samples/tree/master/AI-and-Analytics"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.7.6"
}
},
"nbformat": 4,
"nbformat_minor": 4
}
About
Samples for Intel oneAPI toolkits
Resources
Stars
Watchers
Forks
Packages 0
No packages published
Languages
- Jupyter Notebook 79.7%
- Python 20.1%
- Shell 0.2%