mindformers.core¶

MindFormers Core.

`mindformers.core.build_context`	Build context.
`mindformers.core.init_context`	Context initialization for MindSpore. Args: use_parallel (Optional[Union[bool]]): Whether to use distributed training. Default: False. context_config (Optional[Union[dict, ContextConfig]]): Context Config For Running Environment. Default: None. parallel_config (Optional[Union[dict, ParallelContextConfig]]): Parallel Config For Running Environment. Default: None.

mindformers.core.callback¶

`mindformers.core.callback.CheckpointMointor`	Checkpoint Monitor For Save LossScale.
`mindformers.core.callback.MFLossMonitor`	Loss Monitor for classification.
`mindformers.core.callback.ObsMonitor`	Obs Monitor For Local and AICC.
`mindformers.core.callback.SummaryMonitor`	Summary Monitor For AICC and Local.
`mindformers.core.callback.ProfileMonitor`	Profile analysis in training.
`mindformers.core.callback.EvalCallBack`	Evaluate Callback used in training progress.

`mindformers.core.loss.CrossEntropyLoss`	Calculate the cross entropy loss.
`mindformers.core.loss.L1Loss`	L1Loss for parallel.
`mindformers.core.loss.MSELoss`	Calculate the MSE loss with given logits and labels.
`mindformers.core.loss.SoftTargetCrossEntropy`	Calculate the SoftTargetCrossEntropy loss with given logits and labels.

`mindformers.core.lr.ConstantWarmUpLR`	Constant Warm Up Learning Rate.
`mindformers.core.lr.CosineWithRestartsAndWarmUpLR`	Cosine with Restarts and Warm Up Learning Rate.
`mindformers.core.lr.CosineWithWarmUpLR`	Cosine with Warm Up Learning Rate.
`mindformers.core.lr.LinearWithWarmUpLR`	Linear with Warm Up Learning Rate.
`mindformers.core.lr.PolynomialWithWarmUpLR`	Polynomial with Warm Up Learning Rate.
`mindformers.core.lr.CosineAnnealingLR`	Set the learning rate of each parameter group using a cosine annealing schedule, where \(\eta_{max}\) is set to the initial lr and \(T_{cur}\) is the number of epochs since the last restart in SGDR:
`mindformers.core.lr.CosineAnnealingWarmRestarts`	Set the learning rate of each parameter group using a cosine annealing schedule, where \(\eta_{max}\) is set to the initial lr, \(T_{cur}\) is the number of epochs since the last restart and \(T_{i}\) is the number of epochs between two warm restarts in SGDR:

`mindformers.core.metric.EntityScore`	Compute the f1, precision and recall score of each entity
`mindformers.core.metric.SQuADMetric`	Compute the f1, precision and recall score of each entity
`mindformers.core.metric.PerplexityMetric`	Compute the loss and PPL of each entity
`mindformers.core.metric.ADGENMetric`	Compute the f1, precision and recall score of each entity

This class is almost same with the mindspore's AdamWeightDecay implements, the