CODEWITHSUNDEEP
datasetartificial-intelligencetorch
AlexDtd
AlexDtd

Reputation: 21

About pytorch model input sample size

I'm newbie in pytorch and AI. I found a repo (https://github.com/kacperbednarz1997/AMC_nets) where author compares various models for automatic modulation classification (AMC).

I get all datasets used (RadioML2016.10a, RadioML2016.10b, RadioML2018.01a and MIGOU).

For example, I trained AWN using RadioML2016.10a (each input is IQ, so 2x128 values each). Results for validation and test seems to be good.

I wish to use RadioML2018.01a to test trained model on samples never seen. But, this dataset datas is 2x1024 instead of 2x128. What can I do?

  1. Torch is enough flexible?

    a) 1 input of 1024 is internally split in 8*128 with a repeated label 8 times?

    b) only first 128 taken

  2. I have to prepare my dataset splitting myself as in hypothesis (a) before inference?

Same question if it's inverse configuration: Training with 2x1024 and inference with 2x128.

Thanks for your help

Upvotes: 0

Views: 44

Answers (1)

CNLiu
CNLiu

Reputation: 1

If you want to use a trained model for inference, the input data must conform to the format (including size) in which the model was originally trained. Basically, I think both of your proposed solutions, 1.a) and 1.b), can be tried and tested, and as you mentioned, the data must be processed by yourself first.

However, because you are performing inference on untrained data, and the data format is different from the model training data, I think the model performance will not be as expected.

Please refer to it, thanks.

Upvotes: 0

Related Questions