Skip to content

Latest commit

 

History

History
100 lines (76 loc) · 4.08 KB

README.md

File metadata and controls

100 lines (76 loc) · 4.08 KB
Raw

SSD-ResNet34 Training

Description

This document has instructions for running SSD-ResNet34 training using Intel-optimized PyTorch.

Quick Start Scripts

Script name Description
throughput.sh Tests the training performance for SSD-ResNet34 for the specified precision (fp32, avx-fp32, bf32 or bf16).
accuracy.sh Tests the training accuracy for SSD-ResNet34 for the specified precision (fp32, avx-fp32, bf32 or bf16).

Note: The avx-fp32 precision runs the same scripts as fp32, except that the DNNL_MAX_CPU_ISA environment variable is unset. The environment variable is otherwise set to DNNL_MAX_CPU_ISA=AVX512_CORE_AMX.

  • Set ENV to use AMX: 
    export DNNL_MAX_CPU_ISA=AVX512_CORE_AMX

| Distributed Training |

DataType Throughput Accuracy
FP32 bash throughput_dist.sh fp32 bash accuracy_dist.sh fp32
BF16 bash throughput_dist.sh bf16 bash accuracy_dist.sh bf16
BF32 bash throughput_dist.sh bf32 bash accuracy_dist.sh bf32

Bare Metal

General setup

Follow link to install Miniconda and build Pytorch, IPEX, TorchVison, Torch-CCL TCmalloc and Jemalloc.

Model Specific Setup

  • Set Jemalloc and tcmalloc Preload for better performance

    The jemalloc should be built from the General setup section.

    export LD_PRELOAD="<path to the jemalloc directory>/lib/libjemalloc.so":"path_to/tcmalloc/lib/libtcmalloc.so":$LD_PRELOAD
export MALLOC_CONF="oversize_threshold:1,background_thread:true,metadata_thp:auto,dirty_decay_ms:9000000000,muzzy_decay_ms:9000000000"

  • Set IOMP preload for better performance

    IOMP should be installed in your conda/virtual env from the General setup section.

    pip install packaging intel-openmp
export LD_PRELOAD=path/lib/libiomp5.so:$LD_PRELOAD

  • Set ENV to use multi-node distributed training (no need for single-node multi-sockets)

    In this case, we use data-parallel distributed training and every rank will hold same model replica. The NNODES is the number of ip in the HOSTFILE. To use multi-nodes distributed training you should firstly setup the passwordless login (you can refer to link) between these nodes.

    export NNODES=#your_node_number
export HOSTFILE=your_ip_list_file #one ip per line

Run the model

Once all the setup is done, the Intel® AI Reference Models repo can be used to run a quickstart script. Ensure that you have an enviornment variables set to point to the dataset directory and an output directory.

# Clone the Intel® AI Reference Models riepo and set the MODEL_DIR
git clone https://github.com/IntelAI/models.git
cd models
export MODEL_DIR=$(pwd)

# Dataset
Download the 2017 [COCO dataset](https://cocodataset.org) using the `download_dataset.sh` script.
Export the `DATASET_DIR` environment variable to specify the directory where the dataset
will be downloaded. This environment variable will be used again when running quickstart scripts.

cd quickstart/object_detection/pytorch/ssd-resnet34/training/cpu
export DATASET_DIR=<directory where the dataset will be saved>
bash download_dataset.sh
cd - 
cd ${MODEL_DIR}

# install model specific dependencies
./quickstart/object_detection/pytorch/ssd-resnet34/training/cpu/setup.sh

# Download pretrained model
export CHECKPOINT_DIR=<directory where the pretrained model will be saved>
bash quickstart/object_detection/pytorch/ssd-resnet34/training/cpu/download_model.sh

# Env vars
export DATASET_DIR=<path to the COCO dataset>
export OUTPUT_DIR=<path to an output directory>
export PRECISION=<select from :- fp32, avx-fp32, bf16, or bf32>

# Optional environemnt variables:
export BATCH_SIZE=<set a value for batch size, else it will run with default batch size>

# Run a quickstart script
./quickstart/object_detection/pytorch/ssd-resnet34/training/cpu/<script.sh>

License

LICENSE