TESTING ALL THE READER PIPELINES ON PYTHON IN SINGLE SHOT#
./TEST_FILE.sh
TO TEST A SINGLE READER / MULTIPLE READER PIPELINE FROM TEST_FILE.sh#
Make the respective ” Pipeline ” to test equal to “1”#
rocAL_api_python_unittest=1 rocAL_api_coco_pipeline=0
INSTALL PYBIND TO RUN ANY TESTS#
EXPORT ROCAL_DATA_PATH#
export ROCAL_DATA_PATH=/Absolute/Path/Of/rocal_data/
TESTING ALL THE AUGMENTATIONS IN A SINGLE SHOT#
./PYTHON_UNITTEST_TEST_FILE.sh (The default value of number of gpu’s is “1” & display is “ON” by default)
./PYTHON_UNITTEST_TEST_FILE.sh -n <number_of_gpus> -d <true/false> -b backend<cpu/gpu> -p print_tensor<true/false> ./PYTHON_UNITTEST_TEST_FILE.sh -n “1” -d “false” -b “cpu”
TO TEST MULTIPLE READER PIPELINE FROM READERS_TEST_FILE.sh#
./READERS_TEST_FILE.sh (The default value of number of gpu’s is “1” & display is “ON” by default)
./READERS_TEST_FILE.sh -n <number_of_gpus> -d <true/false> ./READERS_TEST_FILE.sh -n “1” -d “false” -b “cpu”
TO TEST A SINGLE READER PIPELINE FROM READERS_TEST_FILE.sh#
example : To run COCO Pipeline
rocAL_api_python_unittest=0 rocAL_api_coco_pipeline=1 rocAL_api_caffe_reader=0 rocAL_api_caffe2_reader=0 rocAL_api_tf_classification_reader=0 rocAL_api_tf_detection_pipeline=0
TO TEST A SINGLE READER PIPELINE FROM CMD LINE#
example: COCO Pipeline
# Mention the number of gpus
gpus_per_node=4
# Mention Batch Size
batch_size=10
# python version
ver=$(python -V 2>&1 | sed 's/.* \([0-9]\).\([0-9]\).*/\1\.\2/')
# Mention dataset_path
data_dir=$ROCAL_DATA_PATH/coco/coco_10_img/val_10images_2017/
# Mention json path
json_path=$ROCAL_DATA_PATH/coco/coco_10_img/annotations/instances_val2017.json
# rocAL_api_coco_pipeline.py
# By default : cpu backend, NCHW format , fp32
# Annotation must be a json file
python$ver rocAL_api_coco_pipeline.py --image-dataset-path $data_dir --json-path $json_path --batch-size $batch_size --display --rocal-gpu --NHWC \
--local-rank 0 --world-size $gpus_per_node --num-threads 1 --num-epochs 1 2>&1 | tee -a run.log.rocAL_api_coco_pipeline.txt