list indices must be integers or slices, not str

[huutuan1705]

Hello,

I see your code in the internet, and this is so interesting. I cloned it and used for my project. I try changing batch_size from 1 to 4 and backbone from vgg16 to resnet101. But I have a problem. You can see this image to know about my error:

I think the cause of this issue is that I changed the batch_size, which led to changes in the list of boxes. I try to fix this problem for three days, but not success. I really need your help. Can you help me fix that. Here is my code:

https://colab.research.google.com/drive/1YpICgRMK7wiIFn85dt753j18NIY3JgfH?usp=sharing

Thank you so much!!!

[explainingai-code]

Hello @huutuan1705 ,

This repo is only made for single batch size. If you want to train with larger than 1 batch size, then you can use gradient accumulation for that purpose. So batch size would still be 1 but network would be updated with accumulated gradients over multiple batches.
Assuming you want to train with 4 batch size, so just set the acc_steps parameter in config to be 4, that would train the network with batch size of 1 but update the model only after accumulating gradient for 4 batches.

https://github.com/explainingai-code/FasterRCNN-PyTorch/blob/main/config/voc.yaml#L37
You can see how this impacts training code here

Do let me know if you run into any sort of issues with this setting.

[huutuan1705]

Thank you so much, I understand that. Thanks for your help.