implemented Custom model for experimentation with AVES models - also …

…changing input tensor from 3D to 2D due to retraining only last class. layers

aves-base-bio.torchaudio.model_config.json

@@ -0,0 +1,53 @@
+{
+    "extractor_mode": "group_norm",
+    "extractor_conv_layer_config": [
+        [
+            512,
+            10,
+            5
+        ],
+        [
+            512,
+            3,
+            2
+        ],
+        [
+            512,
+            3,
+            2
+        ],
+        [
+            512,
+            3,
+            2
+        ],
+        [
+            512,
+            3,
+            2
+        ],
+        [
+            512,
+            2,
+            2
+        ],
+        [
+            512,
+            2,
+            2
+        ]
+    ],
+    "extractor_conv_bias": false,
+    "encoder_embed_dim": 768,
+    "encoder_projection_dropout": 0.1,
+    "encoder_pos_conv_kernel": 128,
+    "encoder_pos_conv_groups": 16,
+    "encoder_num_layers": 12,
+    "encoder_num_heads": 12,
+    "encoder_attention_dropout": 0.1,
+    "encoder_ff_interm_features": 3072,
+    "encoder_ff_interm_dropout": 0.0,
+    "encoder_dropout": 0.1,
+    "encoder_layer_norm_first": false,
+    "encoder_layer_drop": 0.05
+}