cs273b_params.json

{
    "train": {
        "batch_size": 256,
        "optimizer": "sgd",
        "learning_rate": 0.005,
        "momentum": 0.99,
        "patience": 16,
        "clipnorm": 8
    },
    "model": {
        "seq_length": 1000,
        "target_length": 1,

        "activation": "gelu",
        "batch_norm": true,
        "bn_momentum": 0.9,

        "trunk": [
            {
                "name": "conv_block",
                "filters": 128,
                "kernel_size": 17,
                "pool_size": 5
            },
            {
                "name": "conv_block",
                "filters": 144,
                "kernel_size": 9,
                "pool_size": 5
            },
        {
                "name": "conv_tower",
                "filters_init": 162,
                "filters_mult": 1.125,
                "kernel_size": 5,
                "pool_size": 2,
                "repeat": 3
            }
        ],
        "head": [
        {
                "name": "conv_block",
                "filters": 32,
        "kernel_size": 5
            },
        {
        "name": "Cropping1D",
        "cropping": 2
        },
        {
        "name": "dense",
        "units": 4,
        "activation": "softplus"
        }
    ]
    }
}