config.json

{
  "_name_or_path": "kawn_cohere_8b_idefics3_siglib14_384",
  "architectures": [
    "KawnIdefics3ForConditionalGeneration"
  ],
  "ignore_index": -100,
  "image_grid_pinpoints": [
    [
      256,
      256
    ],
    [
      256,
      512
    ],
    [
      256,
      768
    ],
    [
      256,
      1024
    ],
    [
      256,
      1280
    ],
    [
      1280,
      256
    ],
    [
      1024,
      256
    ],
    [
      768,
      256
    ],
    [
      512,
      256
    ],
    [
      512,
      512
    ],
    [
      512,
      768
    ],
    [
      512,
      1024
    ],
    [
      512,
      1280
    ],
    [
      1280,
      512
    ],
    [
      1024,
      512
    ],
    [
      768,
      512
    ],
    [
      768,
      1024
    ],
    [
      1024,
      768
    ],
    [
      1280,
      768
    ],
    [
      768,
      1280
    ],
    [
      768,
      768
    ],
    [
      1024,
      1024
    ],
    [
      1024,
      1280
    ],
    [
      1280,
      1024
    ],
    [
      1280,
      1280
    ]
  ],
  "image_seq_length": 64,
  "image_token_id": 256001,
  "model_type": "kawn_idefics3",
  "scale_factor": 2,
  "text_config": {
    "_name_or_path": "google/gemma-2-2b-it",
    "architectures": [
      "Gemma2ForCausalLM"
    ],
    "attn_logit_softcapping": 50.0,
    "bos_token_id": 2,
    "cache_implementation": "hybrid",
    "eos_token_id": [
      1,
      107
    ],
    "final_logit_softcapping": 30.0,
    "head_dim": 256,
    "hidden_act": "gelu_pytorch_tanh",
    "hidden_activation": "gelu_pytorch_tanh",
    "hidden_size": 2304,
    "intermediate_size": 9216,
    "model_type": "gemma2",
    "num_attention_heads": 8,
    "num_hidden_layers": 26,
    "num_key_value_heads": 4,
    "query_pre_attn_scalar": 256,
    "rms_norm_eps": 1e-06,
    "sliding_window": 4096,
    "torch_dtype": "bfloat16"
  },
  "torch_dtype": "bfloat16",
  "transformers_version": "4.47.1",
  "use_cache": true,
  "vision_config": {
    "_attn_implementation_autoset": true,
    "hidden_size": 768,
    "image_size": 256,
    "intermediate_size": 3072,
    "model_type": "siglip_vision_model",
    "num_attention_heads": 12,
    "num_hidden_layers": 12,
    "patch_size": 16,
    "vision_use_head": false
  },
  "vision_feature_layer": null
}