ENH Adds no batch dim support for AvgPool1d (pytorch#61860)

Summary:
Towards pytorch#60585

Pull Request resolved: pytorch#61860

Reviewed By: albanD

Differential Revision: D29826382

Pulled By: jbschlosser

fbshipit-source-id: 47e12073d866f0604310fc1ff270cde9907e516d

aten/src/ATen/native/Pooling.cpp

@@ -91,20 +91,20 @@ Tensor avg_pool1d(
   if (stride.empty()) {
     stride = kernel_size;
   }
-  checkDim("avg_pool1d", TensorArg(self, "self", 1), 3);
+  checkDimRange("avg_pool1d", TensorArg(self, "self", 1), 2, 4 /* exclusive */);
   check1d("avg_pool1d", "kernel_size", kernel_size);
   check1d("avg_pool1d", "stride", stride);
   check1d("avg_pool1d", "padding", padding);
 
   auto output = at::avg_pool2d(
-      self.unsqueeze(2),
+      self.unsqueeze(-2),
       {1, kernel_size[0]},
       {1, stride[0]},
       {0, padding[0]},
       ceil_mode,
       count_include_pad);
 
-  return output.squeeze(2);
+  return output.squeeze(-2);
 }
 
 Tensor max_pool2d(

torch/nn/modules/pooling.py

@@ -502,8 +502,8 @@ class AvgPool1d(_AvgPoolNd):
         count_include_pad: when True, will include the zero-padding in the averaging calculation
 
     Shape:
-        - Input: :math:`(N, C, L_{in})`
-        - Output: :math:`(N, C, L_{out})`, where
+        - Input: :math:`(N, C, L_{in})` or :math:`(C, L_{in}`.
+        - Output: :math:`(N, C, L_{out})` or :math:`(C, L_{out})`, where
 
           .. math::
               L_{out} = \left\lfloor \frac{L_{in} +

torch/testing/_internal/common_nn.py

@@ -2109,6 +2109,14 @@ def single_batch_reference_fn(input, parameters, module):
         input_size=(2, 3, 6),
         desc='stride_pad',
     ),
+    dict(
+        module_name='AvgPool1d',
+        constructor_args=(2,),
+        cpp_constructor_args='torch::nn::AvgPool1dOptions(2)',
+        input_size=(3, 6),
+        reference_fn=single_batch_reference_fn,
+        desc='no_batch_dim',
+    ),
     dict(
         module_name='AvgPool2d',
         constructor_args=((2, 2),),