Upload model

config.json

@@ -19,7 +19,7 @@
     "dropout": 0.0,
     "layer_norm_eps": 1e-06,
     "nhead": 12,
-    "norm_first": true
+    "norm_first": false
   },
   "encoder_cfg": {
     "enable_nested_tensor": false,
@@ -34,7 +34,7 @@
     "dropout": 0.0,
     "layer_norm_eps": 1e-06,
     "nhead": 12,
-    "norm_first": true
+    "norm_first": false
   },
   "extractor_config": {
     "conv_bias": false,

configuration_wavjepa.py

@@ -19,7 +19,7 @@ class WavJEPAConfig(PretrainedConfig):
         encoder_d_model: int = 768,
         encoder_nhead : int = 12,
         encoder_batch_first = True,
-        encoder_norm_first = True, 
+        encoder_norm_first = False, 
         encoder_bias = True,
         encoder_mlp_ratio = 4.0,
         encoder_dropout = 0.0,
@@ -29,7 +29,7 @@ class WavJEPAConfig(PretrainedConfig):
         decoder_d_model: int = 384,
         decoder_nhead : int = 12,
         decoder_batch_first = True,
-        decoder_norm_first = True, 
+        decoder_norm_first = False, 
         decoder_bias = True,
         decoder_mlp_ratio = 4.0,
         decoder_dropout = 0.0,