config.json

{
  "architectures": [
    "MiniCPMV4_7ForConditionalGeneration"
  ],
  "downsample_mode": "16x",
  "dtype": "float32",
  "image_end_id": null,
  "image_size": 448,
  "image_start_id": null,
  "image_token_id": 248056,
  "insert_layer_id": 1,
  "merge_kernel_size": [
    2,
    2
  ],
  "merger_times": 1,
  "model_type": "minicpmv4_7",
  "mrope_mode": "canvas",
  "newline_id": null,
  "patch_size": 14,
  "slice_end_id": null,
  "slice_start_id": null,
  "text_config": {
    "attention_bias": false,
    "attention_dropout": 0.0,
    "attn_output_gate": true,
    "bos_token_id": null,
    "eos_token_id": null,
    "full_attention_interval": 4,
    "head_dim": 32,
    "hidden_act": "silu",
    "hidden_size": 8,
    "initializer_range": 0.02,
    "layer_types": [
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention"
    ],
    "linear_conv_kernel_dim": 4,
    "linear_key_head_dim": 8,
    "linear_num_key_heads": 2,
    "linear_num_value_heads": 4,
    "linear_value_head_dim": 8,
    "mamba_ssm_dtype": "float32",
    "max_position_embeddings": 262144,
    "model_type": "qwen3_5_moe_text",
    "moe_intermediate_size": 16,
    "mtp_num_hidden_layers": 0,
    "mtp_use_dedicated_embeddings": false,
    "num_attention_heads": 2,
    "num_experts": 8,
    "num_experts_per_tok": 2,
    "num_hidden_layers": 4,
    "num_key_value_heads": 1,
    "output_router_logits": false,
    "pad_token_id": null,
    "partial_rotary_factor": 0.25,
    "rms_norm_eps": 1e-06,
    "rope_parameters": {
      "mrope_interleaved": true,
      "mrope_section": [
        2,
        1,
        1
      ],
      "partial_rotary_factor": 0.25,
      "rope_theta": 10000000,
      "rope_type": "default"
    },
    "router_aux_loss_coef": 0.001,
    "shared_expert_intermediate_size": 16,
    "tie_word_embeddings": true,
    "use_cache": true,
    "vocab_size": 248144
  },
  "tie_word_embeddings": true,
  "transformers_version": "5.18.0.dev0",
  "video_token_id": 248057,
  "vision_config": {
    "attention_dropout": 0.0,
    "hidden_act": "gelu_pytorch_tanh",
    "hidden_size": 16,
    "image_size": 980,
    "insert_layer_id": 1,
    "intermediate_size": 32,
    "layer_norm_eps": 1e-06,
    "model_type": "minicpmv4_7_vision",
    "num_attention_heads": 2,
    "num_channels": 3,
    "num_hidden_layers": 3,
    "patch_size": 14,
    "window_kernel_size": [
      2,
      2
    ]
  }
}