Ayush Sethi

Hey Team, We are using DLIO to simulate Unet3d e.g. ``` mpirun -np 8 python3 dlio_benchmark/main.py workload=unet3d ++workload.workflow.generate_data=False ++workload.workflow.train=True ++workload.dataset.num_files_train=500000 ++workload.workflow.checkpoint=False ++workload.dataset.data_folder=/mnt/disks/100KB-50GB ++workload.dataset.record_length=100000 ++workload.reader.batch_size=1500 ++workload.reader.read_threads=12 ++workload.reader.file_shuffle=seed ++workload.reader.sample_shuffle=seed ++workload.train.epochs=5 ``` When...

Seems like different threads are reading some overlapping files in torch data loader

12

Hey, I am using the below command to do I/O benchmark: ``` mpirun -np 8 ~/.local/bin/dlio_benchmark workload=unet3d ++workload.train.epochs=1 ++workload.workflow.profiling=True ++workload.profiling.profiler=iostat ++workload.profiling.iostat_devices=[md0] ++workload.dataset.data_folder="../unet3d-data-mounted/" ++workload.dataset.num_files_train=5000 ++workload.reader.batch_size=4 ++workload.dataset.record_length=149989507 ++workload.reader.read_threads=10 ++workload.reader.sample_shuffle=off ``` I am...