config.json

{
  "architectures": [
    "Glm4vForConditionalGeneration"
  ],
  "attention_bias": true,
  "attention_dropout": 0.0,
  "eos_token_id": [
    151329,
    151336,
    151338,
    151348
  ],
  "hidden_act": "silu",
  "hidden_size": 64,
  "image_end_token_id": 151340,
  "image_start_token_id": 151339,
  "image_token_id": 151343,
  "initializer_range": 0.02,
  "intermediate_size": 128,
  "max_position_embeddings": 65536,
  "model_type": "glm4v",
  "num_attention_heads": 2,
  "num_hidden_layers": 2,
  "num_key_value_heads": 1,
  "pad_token_id": 151329,
  "partial_rotary_factor": 0.5,
  "rms_norm_eps": 1e-05,
  "rope_scaling": {
    "mrope_section": [
      2,
      2,
      4
    ],
    "rope_type": "default",
    "type": "default"
  },
  "rope_theta": 10000.0,
  "text_config": {
    "architectures": [
      "Glm4vForConditionalGeneration"
    ],
    "attention_bias": true,
    "attention_dropout": 0.0,
    "eos_token_id": [
      151329,
      151336,
      151338,
      151348
    ],
    "hidden_act": "silu",
    "hidden_size": 64,
    "image_token_id": null,
    "initializer_range": 0.02,
    "intermediate_size": 128,
    "max_position_embeddings": 65536,
    "model_type": "glm4v_text",
    "num_attention_heads": 2,
    "num_hidden_layers": 2,
    "num_key_value_heads": 1,
    "pad_token_id": 151329,
    "partial_rotary_factor": 0.5,
    "rms_norm_eps": 1e-05,
    "rope_scaling": {
      "mrope_section": [
        2,
        2,
        4
      ],
      "rope_type": "default",
      "type": "default"
    },
    "rope_theta": 10000.0,
    "tie_word_embeddings": true,
    "torch_dtype": "bfloat16",
    "use_cache": true,
    "video_token_id": null,
    "vocab_size": 151552
  },
  "torch_dtype": "bfloat16",
  "transformers_version": "4.54.0.dev0",
  "use_cache": true,
  "video_end_token_id": 151342,
  "video_start_token_id": 151341,
  "video_token_id": 151344,
  "vision_config": {
    "attention_bias": false,
    "attention_dropout": 0.0,
    "depth": 2,
    "hidden_act": "silu",
    "hidden_dropout_prob": 0.0,
    "hidden_size": 64,
    "image_size": 336,
    "in_channels": 3,
    "initializer_range": 0.02,
    "intermediate_size": 128,
    "model_type": "glm4v",
    "num_heads": 2,
    "out_hidden_size": 64,
    "patch_size": 14,
    "rms_norm_eps": 1e-05,
    "spatial_merge_size": 2,
    "temporal_patch_size": 2
  },
  "vocab_size": 151552
}