INNER CODE UNIT · Python

lr

nianticlabs/simplerecon · options.py:82

    lr: float = 1e-4
    # weight decay
    wd: float = 1e-4

    # number of sanity validation steps before training
    num_sanity_val_steps: int = 0

    # max number of iterations for training
    max_steps: int = 110000

    # batch size
    batch_size: int = 16
    # validation batch size during training
    val_batch_size: int = 16

    # number of GPUs to use for training.
    gpus: int = 2

View source record →