Batch Inference with Multiple Batches (to be executed on Lynxi brain-inspired systems)

Attention

For batch size > 1, use V1 compiler. The default batch size is 1, and supported batch sizes are 2, 4, 7, 14, 28. Smaller models can support up to 28batch. larger models may report the error ‘Models with unequal numbers of batch and n_slice are not supported yet. Please reduce the batchsize number!’ due to the compiler’s underlying cutoff. please reduce the batchsize value!

应用	执行脚本(以2batch为例)
DVS-MNIST	python3 test.py --config clif3fc3dm_itout-b16x1-dvsmnist --checkpoint best_clif3.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2 python3 test.py --config clif5fc2dm_itout-b16x1-dvsmnist --checkpoint best_clif5.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2
DVS-Gesture	python3 test.py --config clif3flif2dg_itout-b16x1-dvsgesture --checkpoint best_clif3.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2 python3 test.py --config clif7fc1dg_itout-b16x1-dvsgesture --checkpoint best_clif7.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2
CIFAR10-DVS	python3 test.py --config clif5fc2cd_itout-b64x1-cifar10dvs --checkpoint best_clif5.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2 python3 test.py --config clif7fc1cd_itout-b64x1-cifar10dvs --checkpoint best_clif7.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2
Jester	python3 test.py --config resnetlif18-itout-b20x4-16-jester --checkpoint latest.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2
Luna16Cls	python3 test.py --config clif3fc3lc_itout-b16x1-luna16cls --checkpoint latest.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2
RGB-Gesture	python3 test.py --config clif3flif2rg_itout-b16x1-rgbgesture --checkpoint latest.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2
ESImagenet	python3 test.py --config resnetlif18-itout-b64x4-esimagenet --checkpoint latest.pth --use_lyngor 1 --use_legacy 0 --v 1 --b 2