o hr@sNdZddlmZddlmZddlmZeeZ ddiZ Gdd d eZ d S) z Llava model configuration)PretrainedConfig)logging)CONFIG_MAPPINGzllava-hf/llava-v1.5-7bzFhttps://huggingface.co/llava-hf/llava-v1.5-7b/resolve/main/config.jsoncs:eZdZdZdZdZ       d fd d ZZS) LlavaConfiga This is the configuration class to store the configuration of a [`LlavaForConditionalGeneration`]. It is used to instantiate an Llava 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 Llava-9B. e.g. [llava-hf/llava-9b](https://huggingface.co/llava-hf/llava-9b) Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: vision_config (`LlavaVisionConfig`, *optional*): Custom vision config or dict text_config (`Union[AutoConfig, dict]`, *optional*): The config object of the text backbone. Can be any of `LlamaConfig` or `MistralConfig`. ignore_index (`int`, *optional*, defaults to -100): The ignore index for the loss function. image_token_index (`int`, *optional*, defaults to 32000): The image token index to encode the image prompt. projector_hidden_act (`str`, *optional*, defaults to `"gelu"`): The activation function used by the multimodal projector. vision_feature_select_strategy (`str`, *optional*, defaults to `"default"`): The feature selection strategy used to select the vision feature from the CLIP backbone. vision_feature_layer (`int`, *optional*, defaults to -2): The index of the layer to select the vision feature. vocab_size (`int`, *optional*, defaults to 32000): Vocabulary size of the Llava model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`~LlavaForConditionalGeneration`] Example: ```python >>> from transformers import LlavaForConditionalGeneration, LlavaConfig, CLIPVisionConfig, LlamaConfig >>> # Initializing a CLIP-vision config >>> vision_config = CLIPVisionConfig() >>> # Initializing a Llama config >>> text_config = LlamaConfig() >>> # Initializing a Llava llava-1.5-7b style configuration >>> configuration = LlavaConfig(vision_config, text_config) >>> # Initializing a model from the llava-1.5-7b style configuration >>> model = LlavaForConditionalGeneration(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```llavaFN}geludefaultc s||_||_||_||_||_||_||_t|jtr4d|vr#|dnd|d<t |dd i||_n|durGt dddddddd d d |_|j|_||_ t|j trrd|vr\|dnd |d<t |dd i||_ |j j|_n |dur|t d |_ t j d i| dS)N model_typeclip_vision_modeliiiPr i)intermediate_size hidden_size patch_size image_sizenum_hidden_layersnum_attention_heads vocab_sizeprojection_dimllama) ignore_indeximage_token_indexprojector_hidden_actvision_feature_select_strategyvision_feature_layerr vision_config isinstancedictr text_configsuper__init__) selfr!r$rrrrr rkwargs __class__rc/var/www/html/ai/venv/lib/python3.10/site-packages/transformers/models/llava/configuration_llava.pyr&Rs>      zLlavaConfig.__init__)NNrr r r r r )__name__ __module__ __qualname____doc__r is_compositionr& __classcell__rrr)r+rs2rN) r/configuration_utilsrutilsrautor get_loggerr,logger#LLAVA_PRETRAINED_CONFIG_ARCHIVE_MAPrrrrr+s