ayolovesu/caption-generation

1{2  "architectures": [3    "PaliGemmaForConditionalGeneration"4  ],5  "bos_token_id": 2,6  "dtype": "bfloat16",7  "eos_token_id": 1,8  "hidden_size": 2048,9  "ignore_index": -100,10  "image_token_index": 257152,11  "model_type": "paligemma",12  "pad_token_id": 0,13  "projection_dim": 2048,14  "text_config": {15    "attention_bias": false,16    "attention_dropout": 0.0,17    "dtype": "bfloat16",18    "head_dim": 256,19    "hidden_act": "gelu_pytorch_tanh",20    "hidden_activation": null,21    "hidden_size": 2048,22    "initializer_range": 0.02,23    "intermediate_size": 16384,24    "max_position_embeddings": 8192,25    "model_type": "gemma",26    "num_attention_heads": 8,27    "num_hidden_layers": 18,28    "num_image_tokens": 256,29    "num_key_value_heads": 1,30    "rms_norm_eps": 1e-06,31    "rope_theta": 10000.0,32    "use_cache": true,33    "vocab_size": 25721634  },35  "transformers_version": "4.57.3",36  "use_cache": false,37  "vision_config": {38    "attention_dropout": 0.0,39    "dtype": "bfloat16",40    "hidden_act": "gelu_pytorch_tanh",41    "hidden_size": 1152,42    "image_size": 224,43    "intermediate_size": 4304,44    "layer_norm_eps": 1e-06,45    "model_type": "siglip_vision_model",46    "num_attention_heads": 16,47    "num_channels": 3,48    "num_hidden_layers": 27,49    "num_image_tokens": 256,50    "patch_size": 14,51    "projection_dim": 2048,52    "projector_hidden_act": "gelu_fast",53    "vision_use_head": false54  }55}56