rhPdZddlmZddlmZddlmZddlmZeje Z GddeZ Gd d eZ Gd d eZgd Zy)zDia model configuration)Optional)PretrainedConfig)rope_config_validation)loggingc~eZdZdZdZ ddedededededed ed ed ed ed edee deffd Z xZ S)DiaEncoderConfiga This is the configuration class to store the configuration of a [`DiaEncoder`]. It is used to instantiate a Dia encoder according to the specified arguments, defining the encoder architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: max_position_embeddings (`int`, *optional*, defaults to 1024): The maximum sequence length that this model might ever be used with. num_hidden_layers (`int`, *optional*, defaults to 12): Number of hidden layers in the Transformer encoder. hidden_size (`int`, *optional*, defaults to 1024): Dimensionality of the encoder layers and the pooler layer. num_attention_heads (`int`, *optional*, defaults to 16): Number of attention heads for each attention layer in the Transformer encoder. num_key_value_heads (`int`, *optional*, defaults to 16): Number of key and value heads for each attention layer in the Transformer encoder. head_dim (`int`, *optional*, defaults to 128): Dimensionality of the attention head. intermediate_size (`int`, *optional*, defaults to 4096): Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder. norm_eps (`float`, *optional*, defaults to 1e-05): The epsilon used by the normalization layers. vocab_size (`int`, *optional*, defaults to 256): Vocabulary size of the Dia model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`DiaModel`]. hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`, `"relu"`, `"swish"` and `"gelu_new"` are supported. rope_theta (`float`, *optional*, defaults to 10000.0): The base period of the RoPE embeddings. rope_scaling (`dict`, *optional*): Dictionary containing the scaling configuration for the RoPE embeddings. NOTE: if you apply new rope type and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value accordingly. Expected contents: `rope_type` (`str`): The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope', 'llama3'], with 'default' being the original RoPE implementation. `factor` (`float`, *optional*): Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings. In most scaling types, a `factor` of x will enable the model to handle sequences of length x * original maximum pre-trained length. `original_max_position_embeddings` (`int`, *optional*): Used with 'dynamic', 'longrope' and 'llama3'. The original max position embeddings used during pretraining. `attention_factor` (`float`, *optional*): Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention computation. If unspecified, it defaults to value recommended by the implementation, using the `factor` field to infer the suggested value. `beta_fast` (`float`, *optional*): Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear ramp function. If unspecified, it defaults to 32. `beta_slow` (`float`, *optional*): Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear ramp function. If unspecified, it defaults to 1. `short_factor` (`List[float]`, *optional*): Only used with 'longrope'. The scaling factor to be applied to short contexts (< `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden size divided by the number of attention heads divided by 2 `long_factor` (`List[float]`, *optional*): Only used with 'longrope'. The scaling factor to be applied to long contexts (< `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden size divided by the number of attention heads divided by 2 `low_freq_factor` (`float`, *optional*): Only used with 'llama3'. Scaling factor applied to low frequency components of the RoPE `high_freq_factor` (`float`, *optional*): Only used with 'llama3'. Scaling factor applied to high frequency components of the RoPE initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. dia_encodermax_position_embeddingsnum_hidden_layers hidden_sizenum_attention_headsnum_key_value_headshead_dimintermediate_sizenorm_eps vocab_size hidden_act rope_theta rope_scalinginitializer_rangec \||_||_||_||_||_||_||_| |_||_| |_ | |_ | |_ |j*d|jvr|jd|jd<t|| |_ t|<di|y)Ntype rope_type)r r r rrrrrrrrrrrsuper__init__)selfr r r rrrrrrrrrrkwargs __class__s |/var/www/html/ai-insurance-compliance-backend/venv/lib/python3.12/site-packages/transformers/models/dia/configuration_dia.pyrzDiaEncoderConfig.__init__gs"(?$!2&!2#6     $#6 $$(    (Vt7H7H-H-1->->v-FD  k *t$!2 "6")  r#r%ih㈵>silu@N{Gz?) __name__ __module__ __qualname____doc__ model_typeintfloatstrrdictr __classcell__r s@r!r r sGRJ(,!##%#%!% #'+#'##!$#### ## ! ## ! ##############tn##!####r"r c)eZdZdZdZ ddedededededed ed ed ed ed ededededededee dede de f(fd Z xZ S)DiaDecoderConfiga This is the configuration class to store the configuration of a [`DiaDecoder`]. It is used to instantiate a Dia decoder according to the specified arguments, defining the decoder architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: max_position_embeddings (`int`, *optional*, defaults to 3072): The maximum sequence length that this model might ever be used with. num_hidden_layers (`int`, *optional*, defaults to 18): Number of hidden layers in the Transformer decoder. hidden_size (`int`, *optional*, defaults to 2048): Dimensionality of the decoder layers and the pooler layer. intermediate_size (`int`, *optional*, defaults to 8192): Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer decoder. num_attention_heads (`int`, *optional*, defaults to 16): Number of attention heads for each attention layer in the Transformer decoder. num_key_value_heads (`int`, *optional*, defaults to 4): Number of key and value heads for each attention layer in the Transformer decoder. head_dim (`int`, *optional*, defaults to 128): Dimensionality of the attention head. cross_num_attention_heads (`int`, *optional*, defaults to 16): Number of attention heads for each cross-attention layer in the Transformer decoder. cross_head_dim (`int`, *optional*, defaults to 128): Dimensionality of the cross-attention head. cross_num_key_value_heads (`int`, *optional*, defaults to 16): Number of key and value heads for each cross-attention layer in the Transformer decoder. cross_hidden_size (`int`, *optional*, defaults to 1024): Dimensionality of the cross-attention layers. norm_eps (`float`, *optional*, defaults to 1e-05): The epsilon used by the normalization layers. vocab_size (`int`, *optional*, defaults to 1028): Vocabulary size of the Dia model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`DiaModel`]. hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): The non-linear activation function (function or string) in the decoder. If string, `"gelu"`, `"relu"`, `"swish"` and `"gelu_new"` are supported. num_channels (`int`, *optional*, defaults to 9): Number of channels for the Dia decoder. rope_theta (`float`, *optional*, defaults to 10000.0): The base period of the RoPE embeddings. rope_scaling (`dict`, *optional*): Dictionary containing the scaling configuration for the RoPE embeddings. NOTE: if you apply new rope type and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value accordingly. Expected contents: `rope_type` (`str`): The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope', 'llama3'], with 'default' being the original RoPE implementation. `factor` (`float`, *optional*): Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings. In most scaling types, a `factor` of x will enable the model to handle sequences of length x * original maximum pre-trained length. `original_max_position_embeddings` (`int`, *optional*): Used with 'dynamic', 'longrope' and 'llama3'. The original max position embeddings used during pretraining. `attention_factor` (`float`, *optional*): Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention computation. If unspecified, it defaults to value recommended by the implementation, using the `factor` field to infer the suggested value. `beta_fast` (`float`, *optional*): Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear ramp function. If unspecified, it defaults to 32. `beta_slow` (`float`, *optional*): Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear ramp function. If unspecified, it defaults to 1. `short_factor` (`List[float]`, *optional*): Only used with 'longrope'. The scaling factor to be applied to short contexts (< `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden size divided by the number of attention heads divided by 2 `long_factor` (`List[float]`, *optional*): Only used with 'longrope'. The scaling factor to be applied to long contexts (< `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden size divided by the number of attention heads divided by 2 `low_freq_factor` (`float`, *optional*): Only used with 'llama3'. Scaling factor applied to low frequency components of the RoPE `high_freq_factor` (`float`, *optional*): Only used with 'llama3'. Scaling factor applied to high frequency components of the RoPE initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. use_cache (`bool`, *optional*, defaults to `True`): Whether or not the model should return the last key/values attentions (not used by all models). is_encoder_decoder (`bool`, *optional*, defaults to `True`): Indicating that this model is part of an encoder-decoder architecture. dia_decoderr r r rrrrcross_num_attention_headscross_head_dimcross_num_key_value_headscross_hidden_sizerrr num_channelsrrr use_cacheis_encoder_decoderc ||_||_||_||_||_||_||_| |_||_| |_ | |_ | |_ | |_ ||_ ||_||_||_|j *d|j vr|j d|j d<t#|||_||_t)|Tdd|i|y)Nrrr@r)r r r rrrrr<r:r;r=rrrr>rrrrr?rr)rr r r rrrrr:r;r<r=rrrr>rrrr?r@rr s r!rzDiaDecoderConfig.__init__s0(?$!2&!2#6 #6   )B&)B&,!2  $$($(    (Vt7H7H-H-1->->v-FD  k *t$!2" I,>I&Ir")i ii r%r&r%r&r%r#r'ir) r*Nr+TT) r,r-r.r/r0r1r2r3rr4boolrr5r6s@r!r8r8s9UnJ(,!#!%#%#$)+!)+!% #'+#'#'+0J!$0J0J 0J  0J ! 0J!0J0J$'0J0J$'0J0J0J0J0J !0J"#0J$tn%0J&!'0J()0J*!+0J0Jr"r8ceZdZdZdZdgZeedZ dde ede ede de d e d e d e d e e e d e de ffd ZddZxZS) DiaConfigaw This is the configuration class to store the configuration of a [`DiaModel`]. It is used to instantiate a Dia model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the [nari-labs/Dia-1.6B](https://huggingface.co/nari-labs/Dia-1.6B) architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: encoder_config (`DiaEncoderConfig`, *optional*): Configuration for the encoder part of the model. If not provided, a default `DiaEncoderConfig` will be used. decoder_config (`DiaDecoderConfig`, *optional*): Configuration for the decoder part of the model. If not provided, a default `DiaDecoderConfig` will be used. norm_eps (`float`, *optional*, defaults to 1e-05): The epsilon used by the normalization layers. is_encoder_decoder (`bool`, *optional*, defaults to `True`): Indicating that this model uses an encoder-decoder architecture. pad_token_id (`int`, *optional*, defaults to 1025): Padding token id. eos_token_id (`int`, *optional*, defaults to 1024): End of stream token id. bos_token_id (`int`, *optional*, defaults to 1026): Beginning of stream token id. delay_pattern (`list[int]`, *optional*, defaults to `[0, 8, 9, 10, 11, 12, 13, 14, 15]`): The delay pattern for the decoder. The length of this list must match `decoder_config.num_channels`. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. use_cache (`bool`, *optional*, defaults to `True`): Whether or not the model should return the last key/values attentions (not used by all models). Example: ```python >>> from transformers import DiaConfig, DiaModel >>> # Initializing a DiaConfig with default values >>> configuration = DiaConfig() >>> # Initializing a DiaModel (with random weights) from the configuration >>> model = DiaModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ``` diapast_key_values)encoder_configdecoder_configrJrKrr@ pad_token_id eos_token_id bos_token_id delay_patternrr?c  t|tr tdi|}t|tr tdi|}||n t|_||n t|_||_||ngd|_| |_| |_ |j jt|jk(sJdt |4d||||d| y)N) rrD r$ z3Number of channels must match delay pattern length.)rLrMrNr@r) isinstancer4r r8rJrKrrOrr?r>lenrr) rrJrKrr@rLrMrNrOrr?rr s r!rzDiaConfig.__init__Ns nd +-??N nd +-??N0>0JnP`Pb0>0JnP`Pb  .;.G]Mn!2"""//3t7I7I3JJ A J  %%%1    r"c|jS)z^Defaulting to audio config as it's the decoder in this case which is usually the text backbone)rK)rdecoders r!get_text_configzDiaConfig.get_text_configss"""r") NNr'Tir#iNr+T)F)r,r-r.r/r0keys_to_ignore_at_inferencer r8 sub_configsrr2rEr1listrr[r5r6s@r!rGrGs-^J#4"5%5IYZK6:59#'   -1#'#  !12# !!12#  # ! #  # # #  S *# !# # J#r"rG)rGr r8N)r/typingrconfiguration_utilsrmodeling_rope_utilsrutilsr get_loggerr,loggerr r8rG__all__rr"r!rfsd39   H %o#'o#dJJ'JJZ[# [#| @r"