config.json

{
  "architectures": [
    "Kimi_K25VisionModel"
  ],
  "dtype": "float32",
  "hidden_act": "silu",
  "hidden_size": 32,
  "intermediate_size": 32,
  "max_position_embeddings": 512,
  "merge_kernel_size": [
    1,
    1
  ],
  "model_type": "kimi_k25_vision",
  "num_attention_heads": 2,
  "num_hidden_layers": 2,
  "patch_size": 8,
  "pos_emb_height": 4,
  "pos_emb_time": 1,
  "pos_emb_width": 4,
  "rope_parameters": {
    "rope_theta": 10000,
    "rope_type": "default"
  },
  "transformers_version": "5.16.0.dev0"
}