loss_scaler.py

# Copyright (c) 2022, NVIDIA CORPORATION. All rights reserved.

"""For backward compatibility, we need the class definitions to deserialize."""

class LossScaler:
    def __init__(self, scale=1):
        self.cur_scale = scale

class DynamicLossScaler:
    def __init__(self,
                 init_scale=2**32,
                 scale_factor=2.,
                 scale_window=1000,
                 min_scale=1,
                 delayed_shift=1,
                 consecutive_hysteresis=False):
        self.cur_scale = init_scale
        self.cur_iter = 0
        self.last_overflow_iter = -1
        self.scale_factor = scale_factor
        self.scale_window = scale_window
        self.min_scale = min_scale
        self.delayed_shift = delayed_shift
        self.cur_hysteresis = delayed_shift
        self.consecutive_hysteresis = consecutive_hysteresis