Upload AutoEncoder1d

autoencoder.py

@@ -0,0 +1,31 @@
+from transformers import PreTrainedModel
+from audio_encoders_pytorch import AutoEncoder1d as AE1d, TanhBottleneck
+from .autoencoder_config import AutoEncoder1dConfig
+
+bottleneck = { 'tanh': TanhBottleneck }
+
+class AutoEncoder1d(PreTrainedModel):
+
+    config_class = AutoEncoder1dConfig
+
+    def __init__(self, config: AutoEncoder1dConfig):
+        super().__init__(config)
+
+        self.autoencoder = AE1d(
+            in_channels = config.in_channels,
+            patch_size = config.patch_size,
+            channels = config.channels, 
+            multipliers = config.multipliers,
+            factors = config.factors, 
+            num_blocks = config.num_blocks, 
+            bottleneck = bottleneck[config.bottleneck]() 
+        )
+
+    def forward(self, *args, **kwargs):
+        return self.autoencoder(*args, **kwargs)
+
+    def encode(self, *args, **kwargs):
+        return self.autoencoder.encode(*args, **kwargs)
+
+    def decode(self, *args, **kwargs):
+        return self.autoencoder.decode(*args, **kwargs)

config.json

@@ -1,6 +1,10 @@
 {
+  "architectures": [
+    "AutoEncoder1d"
+  ],
   "auto_map": {
-    "AutoConfig": "autoencoder_config.AutoEncoder1dConfig"
+    "AutoConfig": "autoencoder_config.AutoEncoder1dConfig",
+    "AutoModel": "autoencoder.AutoEncoder1d"
   },
   "bottleneck": "tanh",
   "channels": 32,
@@ -32,5 +36,6 @@
     8
   ],
   "patch_size": 4,
+  "torch_dtype": "float32",
   "transformers_version": "4.24.0"
 }

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f5038eca135d02345e04b3ce8b44a2e66e709e5d0a0ebb8f2209349d3087ee6
+size 83290921