^j&dZddlmZddlmZddlmZmZeje Z edeGdd eZ edeGd d eZ edeGd d eZ gdZy)zGroupViT model configuration)strict)PreTrainedConfig)auto_docstringloggingznvidia/groupvit-gcc-yfcc) checkpointceZdZUdZdZdZdZeed<dZ eed<dZ eed <d Z eed <d Z eed <dZ eed<dZeed<dZeed<dZeezed<dZeezed<dZeed<dZeed<dZedzed<dZedzed<d Zeeezdzed!<y)"GroupViTTextConfigaz Example: ```python >>> from transformers import GroupViTTextConfig, GroupViTTextModel >>> # Initializing a GroupViTTextModel with nvidia/groupvit-gcc-yfcc style configuration >>> configuration = GroupViTTextConfig() >>> model = GroupViTTextModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```groupvit_text_model text_configi vocab_size hidden_sizeiintermediate_size num_hidden_layersnum_attention_headsMmax_position_embeddings quick_gelu hidden_acth㈵>layer_norm_epsdropoutattention_dropout{Gz?initializer_range?initializer_factorN pad_token_idi bos_token_idi eos_token_id)__name__ __module__ __qualname____doc__ model_typebase_config_keyr int__annotations__rrrrrrstrrfloatrrrr!r#r$r%list~/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/groupvit/configuration_groupvit.pyr r s  'J#OJK!s!s  #%S%"J" NE GUS[%(us{(#u# ## L#* $L#*$+0L#S /D(0r2r ceZdZUdZdZdZdZeed<dZ eed<dZ eed <d Z e ee ed fzed <d Ze ee ed fzed<dZe ee ed fzed<dZeed<dZee eze eefzed<dZee eze eefzed<dZeed<dZeed<dZeed<dZeezed<dZeezed<d Zeed!<d"Zeed#<d"Zeed$<d%Ze eeze eezd fzed&<d'Zy())GroupViTVisionConfiga depths (`list[int]`, *optional*, defaults to [6, 3, 3]): The number of layers in each encoder block. num_group_tokens (`list[int]`, *optional*, defaults to [64, 8, 0]): The number of group tokens for each stage. num_output_groups (`list[int]`, *optional*, defaults to [64, 8, 8]): The number of output groups for each stage, 0 means no group. assign_eps (`float`, *optional*, defaults to `1.0`): Epsilon used in layer norm assign_mlp_ratio (`list[int]`, *optional*, defaults to `[0.5, 4]`): Ratio used to infer hidden size of MLP layers. Example: ```python >>> from transformers import GroupViTVisionConfig, GroupViTVisionModel >>> # Initializing a GroupViTVisionModel with nvidia/groupvit-gcc-yfcc style configuration >>> configuration = GroupViTVisionConfig() >>> model = GroupViTVisionModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```groupvit_vision_model vision_configirirrr)rr.depths)@rnum_group_tokens)r:r;r;num_output_groupsr8r image_size patch_sizer num_channelsgelurrrrrrrrr r! assign_eps)g?rassign_mlp_ratioc|jt|jk7r9tj d|jdt|jyy)zOPart of `@strict`-powered validation. Validates the architecture of the config.z&Manually setting num_hidden_layers to z1, but we expect num_hidden_layers = sum(depth) = N)rsumr9loggerwarning)selfs r3validate_architecturez*GroupViTVisionConfig.validate_architecturersR  ! !S%5 5 NN89O9O8PQ!!$T[[!1 24  6r2N)r&r'r(r)r*r+rr,r-rrr9r0tupler<r=rr?rArBrr.rr/rrrr!rDrErKr1r2r3r5r5?sV 4)J%OK!s!s*3FDIc3h '34>d3i%S/1>5?tCy5c?2?  47Jd3i%S/1746Jd3i%S/16L#J NE GUS[%(us{(#u# ##JDLd53;'% S0@*AALr2r5ceZdZUdZdZeedZdZe e zdze d<dZ e e zdze d<dZ ee d<d Zee d <d Zee d <d Zee d<dZee d<dZee d<fdZxZS)GroupViTConfiga projection_intermediate_dim (`int`, *optional*, defaults to 4096): Dimensionality of intermediate layer of text and vision projection layers. output_segmentation (`bool`, *optional*, defaults to False): Whether or not to return the segmentation logits. groupvit)r r7Nr r7rprojection_dimiprojection_intermediate_dimg/L F@logit_scale_init_valuerrr r!Foutput_segmentationc |ji}tjdnAt|jtr|jj }n |j}|j i}tjdnAt|j tr|j j }n |j }|jdd}|jdd}|t di|j }|jD]B\}}||vs |||k7s|dk7s||vr d|d|d} nd |d } tj| D|j||tdi|j } d | vr3| d jDcic]\}}t||c}}| d <| jD]B\}}||vs |||k7s|dk7s||vr d|d |d} nd |d } tj| D|j| t di||_tdi||_t |4di|ycc}}w)NzS`text_config` is `None`. Initializing the `GroupViTTextConfig` with default values.zW`vision_config` is `None`. initializing the `GroupViTVisionConfig` with default values.text_config_dictvision_config_dicttransformers_version`zp` is found in both `text_config_dict` and `text_config` but with different values. The value `text_config_dict["z"]` will be used instead.zn`text_config_dict` is provided which will be used to initialize `GroupViTTextConfig`. The value `text_config["z"]` will be overridden.id2labelzv` is found in both `vision_config_dict` and `vision_config` but with different values. The value `vision_config_dict["zt`vision_config_dict` is provided which will be used to initialize `GroupViTVisionConfig`. The value `vision_config["r1)r rHinfo isinstancer to_dictr7r5popitemsupdater.super __post_init__) rJkwargsr r7rUrV_text_config_dictkeyvaluemessage_vision_config_dict __class__s r3razGroupViTConfig.__post_init__s    #K KKm n ((*< =**224K**K    %M KKq r **,@ A ..668M ..M "::&8$?#ZZ(rrs#.3,   H %56!1)!17!1H567+777t56e(%e(7e(P Kr2