fix inference bugs

docs/build/html/_sources/tutorials/snemi.rst.txt

@@ -4,7 +4,7 @@ Neuron Segmentation
 This tutorial provides step-by-step guidance for neuron segmentation with SENMI3D benchmark datasets.
 Dense neuron segmentation in electronic microscopy (EM) images belongs to the category of instance segmentation.
 The methodology is to first predict the affinity map of pixels with an encoder-decoder ConvNets and 
-then generate the segmentation map using a segmentation algorithm (e.g. watershed). 
+then generate the segmentation map using a segmentation algorithm (e.g., watershed). 
 
 The evaluation of segmentation results is based on the `Rand Index <https://en.wikipedia.org/wiki/Rand_index>`_
 and `Variation of Information <https://en.wikipedia.org/wiki/Variation_of_information>`_.

docs/build/html/tutorials/snemi.html

@@ -192,7 +192,7 @@ <h1>Neuron Segmentation<a class="headerlink" href="#neuron-segmentation" title="
 <p>This tutorial provides step-by-step guidance for neuron segmentation with SENMI3D benchmark datasets.
 Dense neuron segmentation in electronic microscopy (EM) images belongs to the category of instance segmentation.
 The methodology is to first predict the affinity map of pixels with an encoder-decoder ConvNets and
-then generate the segmentation map using a segmentation algorithm (e.g. watershed).</p>
+then generate the segmentation map using a segmentation algorithm (e.g., watershed).</p>
 <p>The evaluation of segmentation results is based on the <a class="reference external" href="https://en.wikipedia.org/wiki/Rand_index">Rand Index</a>
 and <a class="reference external" href="https://en.wikipedia.org/wiki/Variation_of_information">Variation of Information</a>.</p>
 <div class="admonition note">

docs/source/tutorials/snemi.rst

@@ -4,7 +4,7 @@ Neuron Segmentation
 This tutorial provides step-by-step guidance for neuron segmentation with SENMI3D benchmark datasets.
 Dense neuron segmentation in electronic microscopy (EM) images belongs to the category of instance segmentation.
 The methodology is to first predict the affinity map of pixels with an encoder-decoder ConvNets and 
-then generate the segmentation map using a segmentation algorithm (e.g. watershed). 
+then generate the segmentation map using a segmentation algorithm (e.g., watershed). 
 
 The evaluation of segmentation results is based on the `Rand Index <https://en.wikipedia.org/wiki/Rand_index>`_
 and `Variation of Information <https://en.wikipedia.org/wiki/Variation_of_information>`_.

torch_connectomics/data/dataset/misc.py

@@ -11,7 +11,7 @@ def count_volume(data_sz, vol_sz, stride):
     return 1 + np.ceil((data_sz - vol_sz) / stride.astype(float)).astype(int)
 
 def crop_volume(data, sz, st=(0, 0, 0)):  # C*D*W*H, C=1
-    st = np.array(st).astype(np.uint8)
+    st = np.array(st).astype(int)
     return data[st[0]:st[0]+sz[0], st[1]:st[1]+sz[1], st[2]:st[2]+sz[2]]
 
 def crop_volume_mul(data, sz, st=(0, 0, 0)):  # C*D*W*H, for multi-channel input