| { | |
| "architectures": [ | |
| "WavJEPAModel" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_wavjepa.WavJEPAConfig", | |
| "AutoModel": "modeling_wavjepa.WavJEPAModel" | |
| }, | |
| "decoder_cfg": { | |
| "enable_nested_tensor": false, | |
| "mask_check": true, | |
| "num_layers": 12 | |
| }, | |
| "decoder_layers_cfg": { | |
| "batch_first": true, | |
| "bias": true, | |
| "d_model": 384, | |
| "dim_feedforward": 1536, | |
| "dropout": 0.0, | |
| "layer_norm_eps": 1e-06, | |
| "nhead": 12, | |
| "norm_first": false | |
| }, | |
| "encoder_cfg": { | |
| "enable_nested_tensor": false, | |
| "mask_check": true, | |
| "num_layers": 12 | |
| }, | |
| "encoder_layers_cfg": { | |
| "batch_first": true, | |
| "bias": true, | |
| "d_model": 768, | |
| "dim_feedforward": 3072, | |
| "dropout": 0.0, | |
| "layer_norm_eps": 1e-06, | |
| "nhead": 12, | |
| "norm_first": false | |
| }, | |
| "extractor_config": { | |
| "conv_bias": false, | |
| "conv_layers_spec": "[(512, 10, 5)] + [(512, 3, 2)] * 4 + [(512,2,2)]", | |
| "depthwise": false, | |
| "dropout": 0.0, | |
| "in_channels": 1, | |
| "mode": "default" | |
| }, | |
| "model_type": "wavjepa-base", | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.53.2" | |
| } | |