config.json

{
  "architectures": [
    "TvltForPreTraining"
  ],
  "attention_probs_dropout_prob": 0.0,
  "audio_mask_ratio": 0.15,
  "audio_mask_type": "frame-level",
  "audio_patch_size": [
    2,
    2
  ],
  "decoder_hidden_size": 64,
  "decoder_intermediate_size": 128,
  "decoder_num_attention_heads": 4,
  "decoder_num_hidden_layers": 2,
  "frequency_length": 16,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 128,
  "id2label": {
    "0": "LABEL_0"
  },
  "image_mask_ratio": 0.75,
  "image_patch_size": [
    2,
    2
  ],
  "image_size": 32,
  "initializer_range": 0.02,
  "intermediate_size": 128,
  "label2id": {
    "LABEL_0": 0
  },
  "layer_norm_eps": 1e-12,
  "loss_type": "classification",
  "model_type": "tvlt",
  "num_attention_heads": 4,
  "num_audio_channels": 1,
  "num_frames": 2,
  "num_hidden_layers": 12,
  "num_image_channels": 3,
  "pixel_mask_ratio": 0.75,
  "qkv_bias": true,
  "spectrogram_length": 32,
  "task_mae": true,
  "task_matching": true,
  "torch_dtype": "float32",
  "transformers_version": "4.28.0.dev0",
  "use_mean_pooling": true
}