-
Hello,
I see you've been using boolean masks to mask out the weights of the pytorch network. Is there a way to use sparse tensors to achieve an actual speed up in inference in Pytorch currently ?
…
-
**Describe the bug**
It seems like any MinkowskiConvolution with stride > 1 produces non-deterministic features when executed on the GPU and no shared coordinate manager is used.
Running on the CP…
-
Hi. Thanks for the codes and the detailed instruction.
I implemented sparse convolution into my encoder:
```python
with tf.variable_scope('featureEncoder'):
auxiShape = (self.inputShape[0], sel…
-
I think we can produce a significant speedup in our execution of LCA. Since most of the neurons in any LCA computation are below threshold for most of the display period, perhaps we could just ignore…
-
When I was creating the TensorRT plugin using libspconv, I encountered that cached_get_nvrtc_params got called. I know the reason is that deps.is_nvrtc is true (in ConvTunerSimple_tune_and_cache.cc). …
-
Hey, thanks for the implementation - works well, also the example. My Question is, is there a speed gain vs the common Conv2d?
-
Hi,
I found your work on involution very interesting, and it relates to other ideas I am working on like deformable convolutions. So I tried reimplementing your idea for sparse point clouds using …
-
-
http://www.matthewzeiler.com/pubs/cvpr2010/cvpr2010.pdf
Building robust low and mid-level image representations, beyond edge primitives, is a long-standing goal in vision. Many existing feature det…
leo-p updated
7 years ago
-
hi,thanks for your awsome work, i wonder that do you have try export voxelnet model onnx? or could you give some suggestions?