Amp

Auto mixed precision api.

class mindnlp.common.amp.DynamicLossScaler(scale_value, scale_factor, scale_window)[源代码]

基类：LossScaler

Dynamic LossScaler

adjust(grads_finite)[源代码]: adjust scale value.

scale(inputs)[源代码]: scale inputs tensor.

unscale(inputs)[源代码]: unscale inputs tensor.

class mindnlp.common.amp.LossScaler(scale_value)[源代码]

基类：object

Basic LossScaler.

adjust(grads_finite)[源代码]: adjust scale value.

scale(inputs)[源代码]: scale inputs tensor.

unscale(inputs)[源代码]: unscale inputs tensor.

class mindnlp.common.amp.NoLossScaler[源代码]

基类：LossScaler

No LossScaler

adjust(grads_finite)[源代码]: adjust scale value.

scale(inputs)[源代码]: scale inputs tensor.

unscale(inputs)[源代码]: unscale inputs tensor.

class mindnlp.common.amp.StaticLossScaler(scale_value)[源代码]

基类：LossScaler

Static LossScaler.

adjust(grads_finite)[源代码]: adjust scale value.

scale(inputs)[源代码]: scale inputs tensor.

unscale(inputs)[源代码]: unscale inputs tensor.

mindnlp.common.amp.all_finite(inputs)[源代码]: whether all inputs tensor are finite.

mindnlp.common.amp.auto_black_list(network, black_list=None)[源代码]: auto cast based on black list

mindnlp.common.amp.auto_mixed_precision(network, amp_level='O1')[源代码]: auto mixed precision cast.

mindnlp.common.amp.auto_white_list(network, white_list=None)[源代码]: auto cast based on white list