config.json

{
  "architectures": [
    "PixtralVisionModel"
  ],
  "attention_dropout": 0.1,
  "dropout": 0.1,
  "dtype": "float32",
  "head_dim": 8,
  "hidden_act": "gelu",
  "hidden_size": 32,
  "image_size": 30,
  "initializer_range": 0.02,
  "intermediate_size": 37,
  "model_type": "pixtral",
  "num_attention_heads": 4,
  "num_channels": 3,
  "num_hidden_layers": 2,
  "patch_size": 2,
  "projection_dim": 32,
  "rope_parameters": {
    "rope_theta": 10000.0,
    "rope_type": "default"
  },
  "transformers_version": "5.16.0.dev0"
}