Changing batch_size causes error in snn_evaluate

[rashmi-ar]

Hello author,

changing batch_size to default 1 for training and testing is causing error while transforming ann to snn (for yolov3-tiny-mp2conv-mp1none-lk2relu-up2tconv.cfg). The following error occurs in snn_evaluate,
File "ann_to_snn.py", line 183, in
snn_results, maps, firing_ratios = snn_evaluate(opt,
File "/PyTorch-Spiking-YOLOv3/spiking_utils/snn_evaluate.py", line 276, in snn_evaluate
output_snn1, output_snn2 = snn(data) # two branches
ValueError: too many values to unpack (expected 2)

Please let me know what may have gone wrong here.
Thanks in advance.

[hanbaobao950123]

maybe #4 can help you.

[sahenderson10]

Hi Rashmi, did you get this code to work? How did you grab the dataset?

[rashmi-ar]

Hi @sahenderson10 ! You can check data/get_coco_dataset.sh file to download images and metadata or download the dataset from https://pjreddie.com/media/files/train2014.zip and https://pjreddie.com/media/files/val2014.zip

[sahenderson10]

I have downloaded val2014 and train2014 and put them in the data file. How do I create the coco.data file to run the code?

[rashmi-ar]

Hi @sahenderson10 ! This may help - #50 (comment)

[rashmi-ar]

Downgrading the python and pytorch version, python=3.7 and pytorch=1.3, solved the issue.