o %ݫi@sTdZddlmZddlmZddlZddlZeGdddZeGdddZ dS) apConfiguration and utility classes for classes for Dynamic Chunk Training, as often used for the training of streaming-capable models in speech recognition. The definition of Dynamic Chunk Training is based on that of the following paper, though a lot of the literature refers to the same definition: https://arxiv.org/abs/2012.05481 Authors * Sylvain de Langen 2023 ) dataclass)OptionalNc@sNeZdZUdZeed< dZeeed< defddZ deefdd Z dS) DynChunkTrainConfigzDynamic Chunk Training configuration object for use with transformers, often in ASR for streaming. This object may be used both to configure masking at training time and for run-time configuration of DynChunkTrain-ready models. chunk_sizeNleft_context_sizereturncCs |jduS)zlReturns true if the left context is infinite (i.e. any chunk can attend to any past frame). N)rselfr \/home/ubuntu/.local/lib/python3.10/site-packages/speechbrain/utils/dynamic_chunk_training.pyis_infinite_left_context+s z,DynChunkTrainConfig.is_infinite_left_contextcCs|jdurdS|j|jS)zReturns the number of left context *frames* (not chunks). If ``None``, the left context is infinite. See also the ``left_context_size`` field. N)rrrr r r left_context_size_frames1s  z,DynChunkTrainConfig.left_context_size_frames) __name__ __module__ __qualname____doc__int__annotations__rrboolr r r r r r rs rc@seZdZUdZeed< eed< eed< eed< eed< eed< dZee ed < dZ ee ed < d d Z d dZ dS) DynChunkTrainConfigRandomSamplera\Helper class to generate a DynChunkTrainConfig at runtime depending on the current stage. Example ------- >>> from speechbrain.core import Stage >>> from speechbrain.utils.dynamic_chunk_training import DynChunkTrainConfig >>> from speechbrain.utils.dynamic_chunk_training import DynChunkTrainConfigRandomSampler >>> # for the purpose of this example, we test a scenario with a 100% >>> # chance of the (24, None) scenario to occur >>> sampler = DynChunkTrainConfigRandomSampler( ... chunkwise_prob=1.0, ... chunk_size_min=24, ... chunk_size_max=24, ... limited_left_context_prob=0.0, ... left_context_chunks_min=16, ... left_context_chunks_max=16, ... test_config=DynChunkTrainConfig(32, 16), ... valid_config=None ... ) >>> one_train_config = sampler(Stage.TRAIN) >>> one_train_config DynChunkTrainConfig(chunk_size=24, left_context_size=None) >>> one_train_config.is_infinite_left_context() True >>> sampler(Stage.TEST) DynChunkTrainConfig(chunk_size=32, left_context_size=16) chunkwise_probchunk_size_minchunk_size_maxlimited_left_context_probleft_context_chunks_minleft_context_chunks_maxN test_config valid_configcCstd|kS)a"Samples a random boolean with a probability, in a way that depends on PyTorch's RNG seed. Arguments --------- prob : float Probability (0..1) to return True (False otherwise). Returns ------- The sampled boolean )torchranditem)r probr r r _sample_bool~s z-DynChunkTrainConfigRandomSampler._sample_boolcCs|tjjjkr7||jr5t|j|j dd }||j r.t|j |j dd }nd}t||SdS|tjjjkrA|jS|tjjjkrK|jStd|)aMIn training stage, samples a random DynChunkTrain configuration. During validation or testing, returns the relevant configuration. Arguments --------- stage : speechbrain.core.Stage Current stage of training or evaluation. In training mode, a random DynChunkTrainConfig will be sampled according to the specified probabilities and ranges. During evaluation, the relevant DynChunkTrainConfig attribute will be picked. Returns ------- The appropriate configuration rrNzUnsupported stage found )sbcoreStageTRAINr$rr randintrrr"rrrrTESTrVALIDrAttributeError)r stagerleft_context_chunksr r r __call__s2   z)DynChunkTrainConfigRandomSampler.__call__) rrrrfloatrrrrrrr$r/r r r r r<s(  r) r dataclassesrtypingrr speechbrainr%rrr r r r s $