config.json

{
  "_name_or_path": "utischoolnlp/ReVision-250M-256-16",
  "architectures": [
    "ReVisionForConditionalGeneration"
  ],
  "hidden_size": 768,
  "image_token_index": 32000,
  "model_type": "revision",
  "projection_dim": 768,
  "sampler_tokens": 64,
  "text_config": {
    "_name_or_path": "OuteAI/Lite-Mistral-150M-v2-Instruct",
    "architectures": [
      "MistralForCausalLM"
    ],
    "head_dim": 48,
    "hidden_size": 768,
    "intermediate_size": 3072,
    "max_position_embeddings": 2048,
    "model_type": "mistral",
    "num_attention_heads": 16,
    "num_hidden_layers": 12,
    "num_image_tokens": 256,
    "torch_dtype": "float32",
    "vocab_size": 32768
  },
  "torch_dtype": "float32",
  "transformers_version": "4.44.0",
  "vision_config": {
    "_name_or_path": "google/siglip-base-patch16-256",
    "image_size": 256,
    "model_type": "siglip_vision_model",
    "projection_dim": 768
  }
}