config.json

{
  "architectures": [
    "ColPaliForRetrieval"
  ],
  "embedding_dim": 128,
  "model_type": "colpali",
  "text_config": {
    "hidden_size": 16,
    "intermediate_size": 16,
    "model_type": "gemma",
    "num_attention_heads": 4,
    "num_hidden_layers": 2,
    "num_image_tokens": 256,
    "num_key_value_heads": 1,
    "vocab_size": 257216
  },
  "torch_dtype": "float32",
  "transformers_version": "4.49.0",
  "vlm_config": {
    "_attn_implementation_autoset": true,
    "_name_or_path": "hf-internal-testing/tiny-random-PaliGemmaForConditionalGeneration",
    "architectures": [
      "PaliGemmaForConditionalGeneration"
    ],
    "bos_token_id": 2,
    "eos_token_id": 1,
    "hidden_size": 16,
    "image_token_index": 257152,
    "model_type": "paligemma",
    "num_image_tokens": 256,
    "pad_token_id": 0,
    "projection_dim": 16,
    "text_config": {
      "hidden_size": 16,
      "intermediate_size": 16,
      "num_attention_heads": 4,
      "num_hidden_layers": 2,
      "num_image_tokens": 256,
      "num_key_value_heads": 1,
      "vocab_size": 257216
    },
    "torch_dtype": "float32",
    "vision_config": {
      "hidden_size": 16,
      "intermediate_size": 32,
      "num_attention_heads": 16,
      "num_hidden_layers": 2,
      "num_image_tokens": 256,
      "patch_size": 14,
      "projection_dim": 16,
      "projector_hidden_act": "gelu_fast",
      "vision_use_head": false
    }
  }
}