| { |
| "architectures": [ |
| "Gemma3nAudioEncoder" |
| ], |
| "conf_attention_chunk_size": 12, |
| "conf_attention_context_left": 13, |
| "conf_attention_context_right": 0, |
| "conf_attention_logit_cap": 50.0, |
| "conf_conv_kernel_size": 5, |
| "conf_num_attention_heads": 8, |
| "conf_num_hidden_layers": 12, |
| "conf_positional_bias_size": 256, |
| "conf_reduction_factor": 4, |
| "conf_residual_weight": 0.5, |
| "gradient_clipping": 10000000000.0, |
| "hidden_size": 1536, |
| "input_feat_size": 128, |
| "model_type": "gemma3n_audio", |
| "rms_norm_eps": 1e-06, |
| "sscp_conv_channel_size": [ |
| 128, |
| 32 |
| ], |
| "sscp_conv_eps": 0.001, |
| "sscp_conv_group_norm_eps": 0.001, |
| "sscp_conv_kernel_size": [ |
| [ |
| 3, |
| 3 |
| ], |
| [ |
| 3, |
| 3 |
| ] |
| ], |
| "sscp_conv_stride_size": [ |
| [ |
| 2, |
| 2 |
| ], |
| [ |
| 2, |
| 2 |
| ] |
| ], |
| "torch_dtype": "float32", |
| "transformers_version": "4.53.0", |
| "vocab_offset": 262272, |
| "vocab_size": 128 |
| } |
|
|
