kf_config.json

{
  "library_name": "kerasformers",
  "kerasformers_version": "1.2.1",
  "model_module": "kerasformers.models.deepseek_vl_hybrid",
  "model_class": "DeepseekVLHybridModel",
  "variant": "deepseek_vl_7b_base",
  "weights": "model.weights.json",
  "schema_version": 2,
  "weight_dtype": "float16",
  "model_type": "deepseek_vl_hybrid",
  "text_config": {
    "vocab_size": 102400,
    "embed_dim": 4096,
    "mlp_dim": 11008,
    "num_layers": 30,
    "num_heads": 32,
    "num_kv_heads": 32,
    "head_dim": 128,
    "norm_eps": 1e-06,
    "rope_theta": 10000.0,
    "tie_embeddings": false
  },
  "vision_config": {
    "embed_dim": 1024,
    "mlp_dim": 4096,
    "num_layers": 24,
    "num_heads": 16,
    "norm_eps": 1e-06,
    "image_size": 384,
    "patch_size": 16
  },
  "high_res_config": {
    "embed_dim": 768,
    "mlp_dim": 3072,
    "num_layers": 12,
    "num_heads": 12,
    "image_size": 1024,
    "patch_size": 16,
    "output_channels": 256,
    "window_size": 14,
    "global_attn_indexes": [
      2,
      5,
      8,
      11
    ],
    "norm_eps": 1e-06
  },
  "image_token_id": 100015
}