config.json

{
  "architectures": [
    "HyperCLOVAXVisionV2ForConditionalGeneration"
  ],
  "dtype": "float32",
  "image_token_id": 3,
  "model_type": "hyperclovax_vision_v2",
  "pad_token_id": 0,
  "text_config": {
    "attention_bias": false,
    "attention_dropout": 0.0,
    "attention_multiplier": 0.25,
    "bos_token_id": 0,
    "embedding_multiplier": 1.0,
    "eos_token_id": 128001,
    "head_dim": 16,
    "hidden_act": "gelu",
    "hidden_size": 32,
    "initializer_range": 0.02,
    "intermediate_size": 32,
    "logits_scaling": 1.0,
    "max_position_embeddings": 512,
    "mlp_bias": false,
    "model_type": "hyperclovax",
    "num_attention_heads": 2,
    "num_hidden_layers": 2,
    "num_key_value_heads": 2,
    "pad_token_id": 0,
    "residual_multiplier": 1.0,
    "rms_norm_eps": 1e-06,
    "rope_parameters": {
      "rope_theta": 10000.0,
      "rope_type": "default"
    },
    "tie_word_embeddings": false,
    "use_cache": true,
    "use_post_norm": true,
    "vocab_size": 128256
  },
  "tie_word_embeddings": false,
  "transformers_version": "5.18.0.dev0",
  "video_token_id": 4,
  "vision_config": {
    "depth": 2,
    "fullatt_block_indexes": [
      7,
      15,
      23,
      31
    ],
    "hidden_act": "gelu",
    "hidden_size": 32,
    "in_channels": 3,
    "initializer_range": 0.02,
    "intermediate_size": 32,
    "model_type": "qwen2_5_vl_vision",
    "num_heads": 4,
    "out_hidden_size": 32,
    "patch_size": 14,
    "rope_parameters": {
      "rope_theta": 10000.0,
      "rope_type": "axial"
    },
    "spatial_merge_size": 1,
    "temporal_patch_size": 2,
    "tokens_per_second": 1,
    "window_size": 112
  }
}