config.json

{
  "architectures": [
    "Gemma4AudioModel"
  ],
  "attention_chunk_size": 4,
  "attention_context_left": 5,
  "attention_context_right": 0,
  "attention_invalid_logits_value": -1000000000.0,
  "attention_logit_cap": 50.0,
  "conv_kernel_size": 3,
  "dtype": "float32",
  "gradient_clipping": 10000000000.0,
  "hidden_act": "silu",
  "hidden_size": 32,
  "initializer_range": 0.02,
  "model_type": "gemma4_audio",
  "num_attention_heads": 4,
  "num_hidden_layers": 2,
  "output_proj_dims": 32,
  "residual_weight": 0.5,
  "rms_norm_eps": 1e-06,
  "subsampling_conv_channels": [
    16,
    8
  ],
  "transformers_version": "5.16.0.dev0",
  "use_clipped_linears": false
}