config.json

{
  "_name_or_path": "distilled-hear-vit-s-canon",
  "architectures": [
    "HearCanonViTModel"
  ],
  "auto_map": {
    "AutoConfig": "configuration_hear_canon.HearCanonViTConfig",
    "AutoModel": "modeling_hear_canon.HearCanonViTModel"
  },
  "clip_seconds": 2.0,
  "canon": true,
  "canon_2d": true,
  "canon_a": true,
  "canon_abcd": true,
  "canon_b": true,
  "canon_b_qkv": false,
  "canon_c": true,
  "canon_causal": false,
  "canon_d": true,
  "canon_kernel": 4,
  "canon_no_pos_enc": true,
  "hidden_act": "gelu",
  "hidden_size": 384,
  "image_size": [
    192,
    128
  ],
  "intermediate_size": 1536,
  "layer_norm_eps": 1e-06,
  "model_type": "hear_canon_vit",
  "num_attention_heads": 6,
  "num_audio_samples": 32000,
  "num_channels": 1,
  "num_hidden_layers": 12,
  "patch_size": 16,
  "pooled_dim": 384,
  "pooler_output_size": 384,
  "sample_rate": 16000,
  "timm_model_name": "vit_small_patch16_224",
  "torch_dtype": "float32",
  "transformers_version": "4.50.3"
}