qi#ddlmZddlmZddlmZmZejeZ GddeZ GddeZ ddgZ y ) )PretrainedConfig)logging)CONFIG_MAPPING AutoConfigcReZdZdZdZddddddZ d fd ZxZS) AudioFlamingo3EncoderConfiga This is the configuration class to store the configuration of an [`AudioFlamingo3Encoder`]. It is used to instantiate an AudioFlamingo3 audio encoder according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the audio encoder of the AudioFlamingo3 architecture. e.g. [nvidia/audio-flamingo-3-hf](https://huggingface.co/nvidia/audio-flamingo-3-hf) Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: num_mel_bins (`int`, *optional*, defaults to 128): Number of mel features used per input features. Should correspond to the value used in the `AudioFlamingo3Processor` class. num_hidden_layers (`int`, *optional*, defaults to 32): Number of encoder layers. num_attention_heads (`int`, *optional*, defaults to 20): Number of attention heads for each attention layer in the Transformer encoder. intermediate_size (`int`, *optional*, defaults to 5120): Dimensionality of the "intermediate" (often named feed-forward) layer in encoder. layerdrop (`float`, *optional*, defaults to 0.0): The LayerDrop probability for the encoder. See the [LayerDrop paper](https://huggingface.co/papers/1909.11556) for more details. activation_function (`str`, *optional*, defaults to `"gelu"`): The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`, `"relu"`, `"silu"` and `"gelu_new"` are supported. hidden_size (`int`, *optional*, defaults to 1280): Dimensionality of the layers. dropout (`float`, *optional*, defaults to 0.0): The dropout probability for all fully connected layers in the embeddings, encoder, and pooler. attention_dropout (`float`, *optional*, defaults to 0.0): The dropout ratio for the attention probabilities. activation_dropout (`float`, *optional*, defaults to 0.0): The dropout ratio for activations inside the fully connected layer. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. scale_embedding (`bool`, *optional*, defaults to `False`): Scale embeddings by dividing by sqrt(hidden_size). max_source_positions (`int`, *optional*, defaults to 1500): The maximum sequence length of log-mel filter-bank features that this model might ever be used with. Example: ```python >>> from transformers import AudioFlamingo3EncoderConfig, AudioFlamingo3Encoder >>> # Initializing an AudioFlamingo3EncoderConfig >>> configuration = AudioFlamingo3EncoderConfig() >>> # Initializing an AudioFlamingo3Encoder (with random weights) >>> model = AudioFlamingo3Encoder(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```audioflamingo3_encoder hidden_sizenum_hidden_layersnum_attention_headsintermediate_size layerdrop)d_modelencoder_layersencoder_attention_headsencoder_ffn_dimencoder_layerdropc t|di|||_||_||_||_||_||_| |_| |_ ||_ | |_ ||_ ||_| |_ | |_y)N)super__init__ num_mel_binsr r r rdropoutattention_dropoutactivation_dropoutactivation_functioninitializer_rangerscale_embeddingmax_source_positions)selfrr r rrrr rrrrrr kwargs __class__s q/opt/pipecat/venv/lib/python3.12/site-packages/transformers/models/audioflamingo3/configuration_audioflamingo3.pyrz$AudioFlamingo3EncoderConfig.__init__\s" "6"(&!2#6 !2 !2"4#6 !2"!2.$8!) igeluir)r)r)g{Gz?Fi)__name__ __module__ __qualname____doc__ model_type attribute_mapr __classcell__r#s@r$r r s[7 r*J!-#8.( M"! 9 9r%r c<eZdZdZdZeedZ dfd ZxZ S)AudioFlamingo3Configa This is the configuration class to store the configuration of an [`AudioFlamingo3ForConditionalGeneration`]. It is used to instantiate an AudioFlamingo3 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 AudioFlamingo3. e.g. [nvidia/audio-flamingo-3-hf](https://huggingface.co/nvidia/audio-flamingo-3-hf) Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: audio_config (`Union[AudioFlamingo3EncoderConfig, dict]`, *optional*, defaults to `AudioFlamingo3EncoderConfig`): The config object or dictionary of the audio backbone. text_config (`Union[AutoConfig, dict]`, *optional*, defaults to `Qwen2Config`): The config object or dictionary of the text backbone. audio_token_id (`int`, *optional*, defaults to 151669): The audio token index to encode the audio prompt. projector_hidden_act (`str`, *optional*, defaults to `"gelu"`): Activation function used in the projector. projector_bias (`bool`, *optional*, defaults to `True`): Whether to include bias terms in the projector. Example: ```python >>> from transformers import AudioFlamingo3ForConditionalGeneration, AudioFlamingo3Config, AudioFlamingo3EncoderConfig, Qwen2Config >>> # Initializing an AudioFlamingo3Encoder config >>> audio_config = AudioFlamingo3EncoderConfig() >>> # Initializing a Qwen2 config >>> text_config = Qwen2Config() >>> # Initializing an AudioFlamingo3 configuration >>> configuration = AudioFlamingo3Config(audio_config, text_config) >>> # Initializing a model from the audioflamingo3 style configuration >>> model = AudioFlamingo3ForConditionalGeneration(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```audioflamingo3) audio_config text_configc ||_t|tr'|jdd|d<t |ddi|}n| t d}||_t|tr'|jdd|d<t |ddi|}n| t d}||_||_||_t|(di|y)Nr/r qwen2r) audio_token_id isinstancedictgetrr6r7projector_hidden_actprojector_biasrr)r!r6r7r:r>r?r"r#s r$rzAudioFlamingo3Config.__init__s- lD ))5)9)9,H`)aL &),|*DEU UL  !)*BCEL( k4 ((3 g(NK %(\)BCRkRK  (13K&$8!, "6"r%)NNiuPr*T) r+r,r-r.r/r r sub_configsrr1r2s@r$r4r4s8) V"J3!K# ##r%r4N) configuration_utilsrutilsrautorr get_loggerr+loggerr r4__all__rr%r$rGsQ 4-   H %d9"2d9NO#+O#d "#@ Ar%