o si3+@sddlZddlZddlmZddlmZddlmZddlm Z m Z m Z ddl m Z mZddlmZmZe d d ZGd d d ejjZGd ddeZe ddZeZdS)N)Optionalseparate) activations) pad_x_to_yscript_if_tracing jitable_shape)cached_download SR_HASHTABLE) is_overriddenmark_deprecatedcCs0|jdkr |dddS|jdkr|dS|S)z0Normalize shape of `x` to [batch, n_chan, time].r)ndimreshape unsqueeze)xrO/home/ubuntu/.local/lib/python3.10/site-packages/asteroid/models/base_models.py_unsqueeze_to_3d s   rcseZdZdZddedeeffdd ZddZe d d Z e j d efd d Z d dZ ddZ ddZddZddZeddZddZddZddZZS) BaseModelaoBase class for serializable models. Defines saving/loading procedures, and separation interface to `separate`. Need to overwrite the `forward` and `get_model_args` methods. Models inheriting from `BaseModel` can be used by :mod:`asteroid.separate` and by the `asteroid-infer` CLI. For models whose `forward` doesn't go from waveform to waveform tensors, overwrite `forward_wav` to return waveform tensors. Args: sample_rate (float): Operating sample rate of the model. in_channels: Number of input channels in the signal. If None, no checks will be performed. r sample_rate in_channelscst||_||_dSN)super__init___BaseModel__sample_rater)selfrr __class__rrr(s  zBaseModel.__init__cOtrNotImplementedErrorrargskwargsrrrforward-szBaseModel.forwardcCs|jS)z+Operating sample rate of the model (float).)rrrrrr0szBaseModel.sample_ratenew_sample_ratecCstdt||_dS)NzoOther sub-components of the model might have a `sample_rate` attribute, be sure to modify them for consistency.)warningswarn UserWarningr)rr)rrrr5s  cOstj|g|Ri|S)z4Convenience for :func:`~asteroid.separate.separate`.rr$rrrr>zBaseModel.separatecOtj|g|Ri|S)z:Convenience for :func:`~asteroid.separate.torch_separate`.)rtorch_separater$rrrr/Br-zBaseModel.torch_separatecOr.)z:Convenience for :func:`~asteroid.separate.numpy_separate`.)rnumpy_separater$rrrr0Fr-zBaseModel.numpy_separatecOr.)z9Convenience for :func:`~asteroid.separate.file_separate`.)r file_separater$rrrr1Jr-zBaseModel.file_separatecOs||g|Ri|S)aSeparation method for waveforms. In case the network's `forward` doesn't have waveforms as input/output, overwrite this method to separate from waveform to waveform. Should return a single torch.Tensor, the separated waveforms. Args: wav (torch.Tensor): waveform array/tensor. Shape: 1D, 2D or 3D tensor, time last. r)rwavr%r&rrr forward_wavNs zBaseModel.forward_wavc Osddlm}t|trt|}tj|dd}n|}d|vr(td |d|vr7td |d |vrFtd ||d  |d |d vrbt|trbt |d |d d <z||d}Wnty|||i|d }Yn w||i|d }| |d|S) aInstantiate separation model from a model config (file or dict). Args: pretrained_model_conf_or_path (Union[dict, str]): model conf as returned by `serialize`, or path to it. Need to contain `model_args` and `state_dict` keys. *args: Positional arguments to be passed to the model. **kwargs: Keyword arguments to be passed to the model. They overwrite the ones in the model package. Returns: nn.Module corresponding to the pretrained model conf/URL. Raises: ValueError if the input config file doesn't contain the keys `model_name`, `model_args` or `state_dict`. r)getcpu) map_location model_namezDExpected config dictionary to have field model_name`. Found only: {} state_dictzDExpected config dictionary to have field state_dict`. Found only: {} model_argszDExpected config dictionary to have field model_args`. Found only: {}rN) r4 isinstancestrr torchloadkeys ValueErrorformatupdater load_state_dict) clspretrained_model_conf_or_pathr%r&r4 cached_modelconf model_classmodelrrrfrom_pretrained[sF          zBaseModel.from_pretrainedcCsVddl}ddlm}t|jj||d}t}ttj|j|d|d<||d<|S) zSerialize model and output dictionary. Returns: dict, serialized model with keys `model_args` and `state_dict`. rNr) __version__)r7r8r9) torch_versionpytorch_lightning_versionasteroid_versionsoftware_versionsinfos) pytorch_lightningr:rKdictr __name__get_state_dictget_model_argsr=)rplrN model_confrPrrr serializes  zBaseModel.serializecCs|S)zEIn case the state dict needs to be modified before sharing the model.)r8r(rrrrTszBaseModel.get_state_dictcCr!)z/Should return args to re-instantiate the class.r"r(rrrrUszBaseModel.get_model_args)r)rS __module__ __qualname____doc__floatrintrr'propertyrsetterrr/r0r1r3 classmethodrJrXrTrU __classcell__rrrrrs$  :rcseZdZdZdfdd ZddZdejdejfd d Zd ejdejfd d Z d ejdejdejfddZ dejdejfddZ ddZ Z S)BaseEncoderMaskerDecoderaiBase class for encoder-masker-decoder separation models. Args: encoder (Encoder): Encoder instance. masker (nn.Module): masker network. decoder (Decoder): Decoder instance. encoder_activation (Optional[str], optional): Activation to apply after encoder. See ``asteroid.masknn.activations`` for valid values. NcsDtjt|ddd||_||_||_||_t|pd|_ dS)Nr)rlinear) rrgetattrencodermaskerdecoderencoder_activationrr4enc_activation)rrerfrgrhrrrrs z!BaseEncoderMaskerDecoder.__init__cCsNt|}t|}||}||}|||}||}t||}t||S)zEnc/Mask/Dec model forward Args: wav (torch.Tensor): waveform tensor. 1D, 2D or 3D tensor, time last. Returns: torch.Tensor, of shape (batch, n_src, time) or (n_src, time). )r rforward_encoderforward_masker apply_masksforward_decoderr_shape_reconstructed)rr2shapetf_rep est_masks masked_tf_repdecoded reconstructedrrrr's       z BaseEncoderMaskerDecoder.forwardr2returncCs||}||S)zComputes time-frequency representation of `wav`. Args: wav (torch.Tensor): waveform tensor in 3D shape, time last. Returns: torch.Tensor, of shape (batch, feat, seq). )reri)rr2rprrrrjs z(BaseEncoderMaskerDecoder.forward_encoderrpcC ||S)zEstimates masks from time-frequency representation. Args: tf_rep (torch.Tensor): Time-frequency representation in (batch, feat, seq). Returns: torch.Tensor: Estimated masks )rf)rrprrrrks z'BaseEncoderMaskerDecoder.forward_maskerrqcCs||dS)a?Applies masks to time-frequency representation. Args: tf_rep (torch.Tensor): Time-frequency representation in (batch, feat, seq) shape. est_masks (torch.Tensor): Estimated masks. Returns: torch.Tensor: Masked time-frequency representations. r)r)rrprqrrrrls z$BaseEncoderMaskerDecoder.apply_masksrrcCrv)zReconstructs time-domain waveforms from masked representations. Args: masked_tf_rep (torch.Tensor): Masked time-frequency representation. Returns: torch.Tensor: Time-domain waveforms. )rg)rrrrrrrms z(BaseEncoderMaskerDecoder.forward_decodercsN|jj|j}tfdd|Dstdi|d|ji}|S)z-Arguments needed to re-instantiate the model.c3s|]}|vVqdSrr).0k fb_configrr sz:BaseEncoderMaskerDecoder.get_model_args..zOFilterbank and Mask network config share common keys. Merging them is not safe.rh)re filterbank get_configrfallAssertionErrorrh)rmasknet_configr9rryrrUs  z'BaseEncoderMaskerDecoder.get_model_argsr)rSrYrZr[rr'r=TensorrjrkrlrmrUrarrrrrbs     rbcCst|dkr |dS|S)zReshape `reconstructed` to have same size as `size` Args: reconstructed (torch.Tensor): Reconstructed waveform size (torch.Tensor): Size of desired waveform Returns: torch.Tensor: Reshaped waveform rr)lensqueeze)rtsizerrrrn$s rn)r=r*typingrr:rmasknnrutils.torch_utilsrrr utils.hub_utilsr r utils.deprecation_utilsr r rnnModulerrbrn BaseTasNetrrrrs      "l