add batch test

extras/generate_test_model.py

@@ -1,7 +1,10 @@
-import torch, torch.nn as nn
-import nn_tilde
 from typing import List, Tuple
 
+import torch
+import torch.nn as nn
+
+import nn_tilde
+
 
 class AudioUtils(nn_tilde.Module):
 
@@ -169,7 +172,7 @@ def get_fractal(self) -> Tuple[int, float]:
     # return -1 if the attribute was wrong.
     @torch.jit.export
     def set_gain_factor(self, x: float) -> int:
-        self.gain_factor= (x,)
+        self.gain_factor = (x, )
         return 0
 
     @torch.jit.export
@@ -182,14 +185,14 @@ def set_polynomial_factors(self, factor1: float, factor2: float,
     @torch.jit.export
     def set_saturate_mode(self, x: str):
         if (x == 'tanh') or (x == 'clip'):
-            self.saturate_mode = (x,)
+            self.saturate_mode = (x, )
             return 0
         else:
             return -1
 
     @torch.jit.export
     def set_invert_signal(self, x: bool):
-        self.invert_signal = (x,)
+        self.invert_signal = (x, )
         return 0
 
     @torch.jit.export
@@ -198,7 +201,7 @@ def set_fractal(self, factor: int, amount: float):
             return -1
         elif factor % 2 != 0:
             return -1
-        self.fractal = ( factor,float(amount))
+        self.fractal = (factor, float(amount))
         return 0
 
 

python_tools/__init__.py

@@ -1,8 +1,8 @@
-from typing import Any, Callable, Optional, Sequence, Tuple, Union
-import torch
-import logging
 import inspect
+import logging
+from typing import Any, Callable, Optional, Sequence, Tuple, Union
 
+import torch
 
 TYPE_HASH = {bool: 0, int: 1, float: 2, str: 3, torch.Tensor: 4}
 
@@ -80,6 +80,9 @@ def register_method(
                 raise ValueError(
                     ("Output tensor must have exactly 3 dimensions, "
                      f"got {len(y.shape)}"))
+            if y.shape[0] != 1:
+                raise ValueError(
+                    f"Expecting single batch output, got {y.shape[0]}")
             if y.shape[1] != out_channels:
                 raise ValueError((
                     f"Wrong number of output channels for method \"{method_name}\", "
@@ -89,7 +92,7 @@ def register_method(
                     (f"Wrong output length for method \"{method_name}\", "
                      f"expected {test_buffer_size//out_ratio} "
                      f"got {y.shape[2]}"))
-            
+
             logging.info(f"Testing method {method_name} with mc.nn~ API")
             x = torch.zeros(4, in_channels, test_buffer_size // in_ratio)
             y = getattr(self, method_name)(x)
@@ -98,6 +101,8 @@ def register_method(
                 raise ValueError(
                     ("Output tensor must have exactly 3 dimensions, "
                      f"got {len(y.shape)}"))
+            if y.shape[0] != 4:
+                raise ValueError(f"Expecting 4 batch output, got {y.shape[0]}")
             if y.shape[1] != out_channels:
                 raise ValueError((
                     f"Wrong number of output channels for method \"{method_name}\", "

setup.py

@@ -21,12 +21,12 @@
     long_description=readme,
     long_description_content_type="text/markdown",
     packages=['nn_tilde'],
-    package_dir= {'nn_tilde': 'python_tools'},
+    package_dir={'nn_tilde': 'python_tools'},
     classifiers=[
         "Programming Language :: Python :: 3",
         "License :: OSI Approved :: MIT License",
         "Operating System :: OS Independent",
     ],
     install_requires=requirements.split("\n"),
     python_requires='>=3.7',
-)
+)