`i/dZddlmZddlmZddlmZmZeje Z GddeZ Gdd eZ Gd d eZ d gZd S) zIdefics2 model configuration)PretrainedConfig)logging)CONFIG_MAPPING AutoConfigcBeZdZdZdZdZ dfd ZxZS)Idefics2VisionConfiga This is the configuration class to store the configuration of a [`Idefics2VisionModel`]. It is used to instantiate a Idefics2 vision 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 SigLIP checkpoint [google/siglip-base-patch16-224](https://huggingface.co/google/siglip-base-patch16-224) used in the Idefics2 model [HuggingFaceM4/idefics2-8b](https://huggingface.co/HuggingFaceM4/idefics2-8b). 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 768): Dimensionality of the encoder layers and the pooler layer. intermediate_size (`int`, *optional*, defaults to 3072): Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder. num_hidden_layers (`int`, *optional*, defaults to 12): Number of hidden layers in the Transformer encoder. num_attention_heads (`int`, *optional*, defaults to 12): Number of attention heads for each attention layer in the Transformer encoder. num_channels (`int`, *optional*, defaults to 3): Number of channels in the input images. image_size (`int`, *optional*, defaults to 224): The size (resolution) of each image. patch_size (`int`, *optional*, defaults to 32): The size (resolution) of each patch. hidden_act (`str` or `function`, *optional*, defaults to `"gelu_pytorch_tanh"`): The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`, `"relu"`, `"selu"` and `"gelu_new"` `"quick_gelu"` are supported. layer_norm_eps (`float`, *optional*, defaults to 1e-06): The epsilon used by the layer normalization layers. attention_dropout (`float`, *optional*, defaults to 0.0): The dropout ratio for the attention probabilities. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation for initializing all weight matrices in the model. Example: ```python >>> from transformers.models.idefics2.modeling_idefics2 import Idefics2VisionTransformer >>> from transformers.models.idefics2.configuration_idefics2 import Idefics2VisionConfig >>> # Initializing a Idefics2VisionConfig with google/siglip-base-patch16-224 style configuration >>> configuration = Idefics2VisionConfig() >>> # Initializing a Idefics2VisionTransformer (with random weights) from the google/siglip-base-patch16-224 style configuration >>> model = Idefics2VisionTransformer(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```idefics2_vision vision_config r gelu_pytorch_tanhư>{Gz?c  tjdi| ||_||_||_||_||_||_||_| |_ | |_ ||_ | |_ dS)N) super__init__ hidden_sizeintermediate_sizenum_hidden_layersnum_attention_heads num_channels patch_size image_sizeattention_dropoutlayer_norm_eps hidden_actinitializer_range)selfrrrrrrrr"r!r r#kwargs __class__s /home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/transformers/models/idefics2/configuration_idefics2.pyrzIdefics2VisionConfig.__init__Os} ""6"""&!2!2#6 ($$!2,$!2) r r rrrrrrrrr)__name__ __module__ __qualname____doc__ model_typebase_config_keyr __classcell__r&s@r'r r sw1 1 f#J%O&3333333333r(r c<eZdZdZdZ dfd ZxZS)Idefics2PerceiverConfigao Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): The non-linear activation function (function or string) in the perceiver block. hidden_size (`int`, *optional*, defaults to 4096): Dimension of the hidden representations. rms_norm_eps (`float`, *optional*, defaults to 1e-06): The epsilon used by the rms normalization layers. resampler_n_latents (`int`, *optional*, defaults to 64): Number of latent embeddings to resample ("compress") the input sequence to (usually < 128). resampler_depth (`int`, *optional*, defaults to 3): Depth of the Perceiver Resampler (Transformer w/ cross attention). Should be shallow (<= 3). resampler_n_heads (`int`, *optional*, defaults to 16): Number of heads in each Transformer block (for multi-headed self-attention). resampler_head_dim (`int`, *optional*, defaults to 96): Dimensionality of each head projection in the Transformer block. num_key_value_heads (`int`, *optional*, defaults to 4): Number of key-value heads in the perceiver attention block. attention_dropout (`float`, *optional*, defaults to 0.0): The dropout ratio for the attention probabilities. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation for initializing all weight matrices in the model. idefics2_perceiversilur@r`rrc  $||_||_||_||_||_||_||_||_| |_| |_ |j|jkrtd|jd|jtj di| dS)Nznum_key_value_heads=z1 must be less than or equal to resampler_n_heads=r) r"r rms_norm_epsresampler_n_latentsresampler_depthresampler_n_headsnum_key_value_headsresampler_head_dimr r# ValueErrorrr) r$r"rr;r<r=r>r@r?r r#r%r&s r'rz Idefics2PerceiverConfig.__init__s%&(#6 .!2#6 "4!2!2  #d&< < <?t'???&*&<??  ""6"""""r() r4r5rr6rr7r8r9rr)r)r*r+r,r-rr/r0s@r'r2r2msn6&J##########r(r2c@eZdZdZdZeeedZ d fd Z xZ S) Idefics2Configa This is the configuration class to store the configuration of a [`Idefics2Model`]. It is used to instantiate a Idefics2 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 model of the Idefics2 [HuggingFaceM4/idefics2-8b](https://huggingface.co/HuggingFaceM4/idefics2-8b) architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: use_cache (`bool`, *optional*, defaults to `True`): Whether or not the model should cache the key/value pairs of the attention mechanism. image_token_id (`int`, *optional*, defaults to 32001): The id of the "image" token. tie_word_embeddings (`bool`, *optional*, defaults to `False`): Whether or not to tie the word embeddings with the token embeddings. vision_config (`IdeficsVisionConfig` or `dict`, *optional*): Custom vision config or dict perceiver_config (`IdeficsPerceiverConfig` or `dict`, *optional*): Custom perceiver config or dict text_config (`MistralConfig` or `dict`, *optional*): Custom text config or dict for the text model Example: ```python >>> from transformers import Idefics2Model, Idefics2Config >>> # Initializing configuration >>> configuration = Idefics2Config() >>> # Initializing a model from the configuration >>> model = Idefics2Model(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```idefics2) text_configperceiver_configr T}FNc ||_||_||_|.t|_t dnCt|trtd i||_nt|tr||_|.t|_ t dnCt|trtd i||_ nt|tr||_ t|tr2| dd|d<t|dd i|}n5|3t dtddddd }||_ |j j|jjkrF|j j|j_|j j|j_t d t#jd i|d |idS)Nz7perciver_config is None, using default perceiver configz2vision_config is None, using default vision configr-mistralz.text_config is None, using default text configigh㈵>F)max_position_embeddingsr; pad_token_idtie_word_embeddingszPerceiver config has a different `hidden_size` than text config, which means default values were used. In your model's config on the hub, add `hidden_size` and `rms_norm_eps` keys under the `perceiver_config` dict. rMr)image_token_id use_cacherMr2rFloggerinfo isinstancedictr r getrrErr; warning_oncerr) r$rOrNrMr rFrEr%r&s r'rzIdefics2Config.__init__s-"#6  #$;$=$=D ! KKQ R R R R ($ / / 5$;$O$O>N$O$OD ! ! (*A B B 5$4D !  !5!7!7D  KKL M M M M  t , , /!5!F!F !F!FD    '; < < /!.D  k4 ( ( (3 i(P(PK %(\)BCRRkRRKK  KKH I I I(3(0!$) K'   '4+@+L L L040@0LD ! -151A1ND ! .   C    KK6KK7JKKKKKKr()TrGFNNN) r)r*r+r,r-rr2r sub_configsrr/r0s@r'rCrCs   DJ!3-K!4L4L4L4L4L4L4L4L4L4Lr(rCN)r,configuration_utilsrutilsrautorr get_loggerr)rPr r2rC__all__rr(r'r\s#"333333--------  H % %R3R3R3R3R3+R3R3R3j;#;#;#;#;#.;#;#;#|^L^L^L^L^L%^L^L^LB  r(