Stonesjtu/Interstellar-CNN-scheduler

Tool for optimize CNN blocking

A new start!

CNN-blocking

Tool for optimize CNN blocking

usage: run_optimizer.py [-h] [-s SCHEDULE] [-v] {basic,mem_explore, dataflow_explore} arch layer

positional arguments:

{basic,mem_explore, dataflow_explore} optimizer type

arch architecture specification

layer layer specification

optional arguments:

-h, --help show this help message and exit

-s SCHEDULE, --schedule SCHEDULE restriction of the schedule space this is optional but restricting the schedule space will accelerate the scipt significantly

-v, --verbose vebosity

Examples

To optimize loop blocking.

Dataflow: Eyeriss

Memory Architecture: 3 level

Network: AlexNet Conv2 Batch16

python run_optimizer.py -v -s ./examples/schedule/eyeriss_alex_conv2.json basic ./examples/arch/3_level_mem_baseline_asic.json ./examples/layer/alex_conv2_batch16.json

Dataflow: TPU

Memory Architecture: 3 level

Network: AlexNet Conv2 Batch16

python run_optimizer.py -v -s ./examples/schedule/tpu.json basic ./examples/arch/3_level_mem_baseline_asic.json ./examples/layer/alex_conv2_batch16.json

To optimize memory capacity.

Dataflow: Eyeriss

Memory Architecture: 3 level

Network: AlexNet Conv2 Batch16

python run_optimizer.py -v -s ./examples/schedule/eyeriss_alex_conv2.json mem_explore ./examples/arch/3_level_mem_explore.json ./examples/layer/alex_conv2_batch16.json

To explore dataflow.

Dataflow: All

Memory Architecture: Eyeriss

Network: AlexNet Conv2 Batch16

python run_optimizer.py -v dataflow_explore ./examples/arch/3_level_mem_baseline_asic.json ./examples/layer/alex_conv2_batch16.json

or:

python ./tools/run_optimizer.py -v -n user_defined_pickle_filename dataflow_explore ./examples/arch/3_level_mem_baseline_asic.json ./examples/layer/alex_conv3_batch16.json