`i'dZddlmZddlmZddlmZejeZ GddeZ Gdd eZ d dgZ d S) zParakeet model configuration.)Union)PretrainedConfig)loggingcVeZdZdZdZdgZ dfd ZxZS)ParakeetEncoderConfiga This is the configuration class to store the configuration of a [`ParakeetEncoder`]. It is used to instantiate a `ParakeetEncoder` model according to the specified arguments, defining the model architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: hidden_size (`int`, *optional*, defaults to 1024): Dimension of the layers and the hidden states. num_hidden_layers (`int`, *optional*, defaults to 24): Number of hidden layers in the Transformer encoder. num_attention_heads (`int`, *optional*, defaults to 8): Number of attention heads for each attention layer in the Transformer encoder. intermediate_size (`int`, *optional*, defaults to 4096): Dimension of the "intermediate" (often named feed-forward) layer in the Transformer encoder. hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): The non-linear activation function (function or string) in the encoder and pooler. attention_bias (`bool`, *optional*, defaults to `True`): Whether to use bias in the attention layers. conv_kernel_size (`int`, *optional*, defaults to 9): The kernel size of the convolution layers in the Conformer block. subsampling_factor (`int`, *optional*, defaults to 8): The factor by which the input sequence is subsampled. subsampling_conv_channels (`int`, *optional*, defaults to 256): The number of channels in the subsampling convolution layers. num_mel_bins (`int`, *optional*, defaults to 80): Number of mel features. subsampling_conv_kernel_size (`int`, *optional*, defaults to 3): The kernel size of the subsampling convolution layers. subsampling_conv_stride (`int`, *optional*, defaults to 2): The stride of the subsampling convolution layers. dropout (`float`, *optional*, defaults to 0.1): The dropout ratio for all fully connected layers in the embeddings, encoder, and pooler. dropout_positions (`float`, *optional*, defaults to 0.0): The dropout ratio for the positions in the input sequence. layerdrop (`float`, *optional*, defaults to 0.1): The dropout ratio for the layers in the encoder. activation_dropout (`float`, *optional*, defaults to 0.1): The dropout ratio for activations inside the fully connected layer. attention_dropout (`float`, *optional*, defaults to 0.1): The dropout ratio for the attention layers. max_position_embeddings (`int`, *optional*, defaults to 5000): The maximum sequence length that this model might ever be used with. scale_input (`bool`, *optional*, defaults to `True`): Whether to scale the input embeddings. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. Example: ```python >>> from transformers import ParakeetEncoderModel, ParakeetEncoderConfig >>> # Initializing a `ParakeetEncoder` configuration >>> configuration = ParakeetEncoderConfig() >>> # Initializing a model from the configuration >>> model = ParakeetEncoderModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ``` This configuration class is based on the ParakeetEncoder architecture from NVIDIA NeMo. You can find more details and pre-trained models at [nvidia/parakeet-ctc-1.1b](https://huggingface.co/nvidia/parakeet-ctc-1.1b). parakeet_encoderpast_key_valuessiluT Pr皙?{Gz?c tjdi|||_||_||_||_||_||_||_|dz dzdkrtd|||_ | |_ | |_ ||_ | |_| |_| |_||_||_||_||_||_||_||_dS)Nrrz"conv_kernel_size must be odd, got )super__init__ hidden_sizenum_hidden_layersnum_attention_headsnum_key_value_headsintermediate_size hidden_actattention_bias ValueErrorconv_kernel_sizesubsampling_conv_kernel_sizesubsampling_conv_stridesubsampling_factorsubsampling_conv_channels num_mel_binsdropoutdropout_positions layerdropactivation_dropoutattention_dropoutmax_position_embeddings scale_inputinitializer_range)selfrrrr!r"r#r%r(r)r*r&r'r+r,r-r.r/r0r1r2kwargs __class__s /home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/transformers/models/parakeet/configuration_parakeet.pyrzParakeetEncoderConfig.__init__as0      '!2#6 #6 !2$, q A % * *TBRTTUU U 0,H)'>$"4)B&( !2""4!2'>$&!2)r r r rrTrr rrrrrrrrrrTr)__name__ __module__ __qualname____doc__ model_typekeys_to_ignore_at_inferencer __classcell__r5s@r6rrsAAF$J#4"5"%%& ! $+53535353535353535353r7rcleZdZdZdZdeiZ d deeefffd Z e defd Z xZ S) ParakeetCTCConfiga This is the configuration class to store the configuration of a [`ParakeetForCTC`]. It is used to instantiate a Parakeet CTC model according to the specified arguments, defining the model architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: vocab_size (`int`, *optional*, defaults to 1025): Vocabulary size of the model. ctc_loss_reduction (`str`, *optional*, defaults to `"mean"`): Specifies the reduction to apply to the output of `torch.nn.CTCLoss`. Only relevant when training an instance of [`ParakeetForCTC`]. ctc_zero_infinity (`bool`, *optional*, defaults to `True`): Whether to zero infinite losses and the associated gradients of `torch.nn.CTCLoss`. Infinite losses mainly occur when the inputs are too short to be aligned to the targets. Only relevant when training an instance of [`ParakeetForCTC`]. encoder_config (`Union[dict, ParakeetEncoderConfig]`, *optional*): The config object or dictionary of the encoder. pad_token_id (`int`, *optional*, defaults to 1024): Padding token id. Also used as blank token id. Example: ```python >>> from transformers import ParakeetForCTC, ParakeetCTCConfig >>> # Initializing a Parakeet configuration >>> configuration = ParakeetCTCConfig() >>> # Initializing a model from the configuration >>> model = ParakeetForCTC(configuration) >>> # Accessing the model configuration >>> configuration = model.config ``` This configuration class is based on the Parakeet CTC architecture from NVIDIA NeMo. You can find more details and pre-trained models at [nvidia/parakeet-ctc-1.1b](https://huggingface.co/nvidia/parakeet-ctc-1.1b). parakeet_ctcencoder_configmeanTNr c ||_||_||_t|trt di||_n|t |_|j|_|jj|_tj dd|i|dS)N pad_token_idr) vocab_sizectc_loss_reductionctc_zero_infinity isinstancedictrrCr2rr)r3rHrIrJrCrGr4r5s r6rzParakeetCTCConfig.__init__s%"4!2 nd + + :"7"I"I."I"ID    #"7"9"9D "1!%!4!F  %      r7c :|dd|i|S)z Instantiate a [`ParakeetCTCConfig`] (or a derived class) from parakeet encoder model configuration. Returns: [`ParakeetCTCConfig`]: An instance of a configuration object rCr)to_dict)clsrCr4s r6from_encoder_configz%ParakeetCTCConfig.from_encoder_configs+sEE."8"8":":EfEEEr7)rDrETNr ) r8r9r:r;r<r sub_configsrrLr classmethodrPr>r?s@r6rArAs&&P J#%:;K!=A   d$99:       4F1FFFF[FFFFFr7rAN) r;typingrconfiguration_utilsrutilsr get_loggerr8loggerrrA__all__rr7r6rYs$#333333  H % %|3|3|3|3|3,|3|3|3~OFOFOFOFOF(OFOFOFd  7 8r7