config.json

{
  "architectures": [
    "InstructBlipForConditionalGeneration"
  ],
  "image_token_index": 32001,
  "initializer_factor": 1.0,
  "initializer_range": 0.02,
  "model_type": "instructblip",
  "num_query_tokens": 32,
  "qformer_config": {
    "model_type": "instructblip_qformer",
    "vocab_size": 30523
  },
  "text_config": {
    "architectures": [
      "LLaMAForCausalLM"
    ],
    "attention_bias": false,
    "bos_token_id": 0,
    "eos_token_id": 1,
    "head_dim": 128,
    "hidden_act": "silu",
    "hidden_size": 5120,
    "initializer_range": 0.02,
    "intermediate_size": 13824,
    "max_sequence_length": 2048,
    "mlp_bias": false,
    "model_type": "llama",
    "num_attention_heads": 40,
    "num_hidden_layers": 40,
    "num_key_value_heads": 40,
    "pad_token_id": 32000,
    "pretraining_tp": 1,
    "rms_norm_eps": 1e-06,
    "rope_scaling": null,
    "rope_theta": 10000.0,
    "tie_word_embeddings": false,
    "torch_dtype": "float16",
    "vocab_size": 32064
  },
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "transformers_version": "4.47.0.dev0",
  "use_decoder_only_language_model": true,
  "video_token_index": 32002,
  "vision_config": {
    "model_type": "instructblip_vision_model"
  }
}