How can I get the models in datasets.zip?

[SakuraiYuuta]

trafficstars

I found the dataset generation method in paper:

For each model (e.g., AlexNet), we generate 2,000 variants (e.g., AlexNets) by re-sampling the output channel number and kernel size for each layer. Specifically, the new output channel number is randomly selected from [0.2 × 𝐶𝑜𝑢𝑡, 1.8 × 𝐶𝑜𝑢𝑡], and the kernel size is sampled from {1, 3, 5, 7, 9}

and the generated models in the datasets.zip, but how can I get the models to execute them by myself?

I noticed that these models are ir graphs, how can I transfer them back to pytorch/tensorflow models to do the inference by myself