{
  "architectures": [
    "Lfm2VlForConditionalGeneration"
  ],
  "downsample_factor": 4,
  "dtype": "float32",
  "image_token_id": 4,
  "model_type": "lfm2_vl",
  "projector_bias": true,
  "projector_hidden_act": "gelu",
  "projector_hidden_size": 32,
  "projector_use_layernorm": true,
  "text_config": {
    "block_auto_adjust_ff_dim": true,
    "block_ffn_dim_multiplier": 1.0,
    "block_multiple_of": 2,
    "bos_token_id": 1,
    "conv_L_cache": 3,
    "conv_bias": false,
    "eos_token_id": 7,
    "full_attn_idxs": [
      0
    ],
    "hidden_size": 32,
    "initializer_range": 0.02,
    "intermediate_size": 37,
    "layer_types": [
      "full_attention",
      "conv"
    ],
    "max_position_embeddings": 200,
    "model_type": "lfm2",
    "norm_eps": 1e-05,
    "num_attention_heads": 4,
    "num_hidden_layers": 2,
    "num_key_value_heads": 2,
    "pad_token_id": 0,
    "rope_parameters": {
      "rope_theta": 1000000.0,
      "rope_type": "default"
    },
    "tie_word_embeddings": true,
    "use_cache": true,
    "vocab_size": 64400
  },
  "tie_word_embeddings": true,
  "transformers_version": "5.16.0.dev0",
  "vision_config": {
    "attention_dropout": 0.0,
    "hidden_act": "gelu_pytorch_tanh",
    "hidden_size": 32,
    "intermediate_size": 37,
    "layer_norm_eps": 1e-06,
    "model_type": "siglip2_vision_model",
    "num_attention_heads": 2,
    "num_channels": 3,
    "num_hidden_layers": 2,
    "num_patches": 16,
    "patch_size": 4
  }
}