hf-tiny-v2/tiny-random-CLIPSegModel
06
1{2 "architectures": [3 "CLIPSegModel"4 ],5 "conditional_layer": 0,6 "decoder_attention_dropout": 0.0,7 "decoder_hidden_act": "quick_gelu",8 "decoder_intermediate_size": 2048,9 "decoder_num_attention_heads": 4,10 "dtype": "float32",11 "extract_layers": [12 113 ],14 "initializer_factor": 1.0,15 "logit_scale_init_value": 2.6592,16 "model_type": "clipseg",17 "projection_dim": 64,18 "reduce_dim": 32,19 "text_config": {20 "attention_dropout": 0.1,21 "bos_token_id": 49406,22 "dropout": 0.1,23 "eos_token_id": 49407,24 "hidden_act": "quick_gelu",25 "hidden_size": 32,26 "initializer_factor": 1.0,27 "initializer_range": 0.02,28 "intermediate_size": 37,29 "layer_norm_eps": 1e-05,30 "max_position_embeddings": 512,31 "model_type": "clipseg_text_model",32 "num_attention_heads": 4,33 "num_hidden_layers": 2,34 "pad_token_id": 49407,35 "vocab_size": 4940836 },37 "transformers_version": "5.16.0.dev0",38 "use_complex_transposed_convolution": false,39 "vision_config": {40 "attention_dropout": 0.1,41 "dropout": 0.1,42 "hidden_act": "quick_gelu",43 "hidden_size": 32,44 "image_size": 30,45 "initializer_factor": 1.0,46 "initializer_range": 0.02,47 "intermediate_size": 37,48 "layer_norm_eps": 1e-05,49 "model_type": "clipseg_vision_model",50 "num_attention_heads": 4,51 "num_channels": 3,52 "num_hidden_layers": 2,53 "patch_size": 254 }55}56 