{
  "architectures": [
    "PerceiverForMaskedLM"
  ],
  "attention_probs_dropout_prob": 0.1,
  "audio_samples_per_frame": 1920,
  "cross_attention_shape_for_attention": "kv",
  "cross_attention_widening_factor": 1,
  "d_latents": 1280,
  "d_model": 768,
  "hidden_act": "gelu",
  "image_size": 56,
  "initializer_range": 0.02,
  "layer_norm_eps": 1e-12,
  "max_position_embeddings": 2048,
  "model_type": "perceiver",
  "num_blocks": 1,
  "num_cross_attention_heads": 8,
  "num_frames": 16,
  "num_hidden_layers": 6,
  "num_latents": 256,
  "num_self_attends_per_block": 26,
  "num_self_attention_heads": 8,
  "output_shape": [
    1,
    16,
    224,
    224
  ],
  "qk_channels": null,
  "samples_per_patch": 16,
  "self_attention_widening_factor": 1,
  "torch_dtype": "float32",
  "train_size": [
    368,
    496
  ],
  "transformers_version": "4.19.4",
  "use_query_residual": true,
  "v_channels": null,
  "vocab_size": 262
}