0.14.6 Release

* Sync backend changes
  * Consistent C function signatures.

* Update changelog and readme

* Minor fix

.gitignore

@@ -133,3 +133,6 @@ dmypy.json
 
 # Output files
 *.out
+
+# Don't commit manifest.in
+MANIFEST.in

CHANGELOG.md

@@ -1,5 +1,16 @@
 # Changelog
 
+## [0.14.6] - 2023-03-21
+Just a really small update that syncs the changes to the private branch.
+It's mostly about the changed signature in both QK and AV (both 1D and 2D), where we now take in both kernel size and dilation.
+Up to 0.14.4 we only took in dilation because QK always took in RPB, and RPB's shape is a function of kernel size, and AV took 
+in attention weights, whose last axis is of size kernel size (kernel size squared in 2D).
+As of 0.14.5, we support optional RPB, which means now we have to take in kernel size in QK. To make the signatures consistent,
+we added kernel size to AV as well, along with assertions that kernel sizes match up, but that was not synced from the private
+branch when we pushed out 0.14.5 because we wanted to support PyTorch 2.0 as soon as possible.
+0.14.6 is just mostly to keep the master branch consistent with the latest release, since the signature difference would create
+an inconsistency between the pip package and the master branch.
+
 ## [0.14.5] - 2023-03-16
 
 ### Added

LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2022 SHI Lab
+Copyright (c) 2023 Ali Hassani.
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

Makefile

@@ -1,9 +1,11 @@
-.PHONY: clean uninstall install test style quality
+.PHONY: sdist clean uninstall install-deps install test style quality
 
 check_dirs := src tests
 
 all: clean uninstall install
 
+full: clean uninstall install-deps install
+
 sdist:
 	@echo "Generating source dist"
 	python3 setup.py sdist
@@ -32,7 +34,6 @@ install-deps:
 
 install: 
 	@echo "Installing NATTEN from source"
-	${MAKE} install-deps
 	pip install -e . 2>&1 | tee install.out
 
 test:

README.md

@@ -50,18 +50,26 @@ We also welcome contributions in all forms.
 Just refer to our website, [shi-labs.com/natten](https://www.shi-labs.com/natten/), select your PyTorch version and the CUDA
 version it was compiled with, copy-paste the command and install in seconds!
 
-For example, if you're on `torch==1.12.1+cu116`, you should install NATTEN using the following wheel:
+For example, if you're on `torch==2.0.0+cu118`, you should install NATTEN using the following wheel:
 ```bash
-pip3 install natten -f https://shi-labs.com/natten/wheels/cu116/torch1.12.1/index.html
+pip3 install natten -f https://shi-labs.com/natten/wheels/cu118/torch2.0.0/index.html
 ```
 
 More generally:
 ```bash
 pip3 install natten -f https://shi-labs.com/natten/wheels/{cu_version}/torch{torch_version}/index.html
 ```
 
-**NOTE:** If you do not specify a wheel URL, you will install a "placeholder" version of NATTEN, which is not usable.
-We strongly recommend using our website, or building from source.
+**NOTE:** If you do not specify a wheel URL, pip will collect NATTEN and try to compile on locally, which depending
+on your system might take up to 30 minutes.
+We strongly recommend using our website if you're a Linux user.
+
+### Mac
+Unfortunately we are not yet able to build Mac wheels, but you can compile on install, so just run:
+
+```bash
+pip3 install natten
+```
 
 ### Windows
 NATTEN should support Windows devices with CUDA, but does not yet have Windows wheels.
@@ -74,13 +82,13 @@ Once you've set up your Python environment and installed PyTorch with CUDA, simp
 pip install ninja # Recommended, not required
 git clone https://github.com/SHI-Labs/NATTEN
 cd NATTEN
-pip install -e .
+make
 ```
 
-#### Optional: unit tests
+#### Optional: run unit tests
 You can optionally run unit tests to verify building from source finished successfully:
 ```bash
-python -m unittest discover -v -s ./tests
+make test
 ```
 
 
@@ -93,7 +101,7 @@ python -m unittest discover -v -s ./tests
 - [ ] Neighborhood Attention 3D (CPU)
 - [x] Dilation support
 - [x] Float16 support and utilization
-- [ ] BFloat16 support (awaiting CUDA 11.8/12 builds of torch)
+- [ ] BFloat16 support
 - [ ] Kepler and Maxwell (30<=SM<60) support
 - [ ] Windows builds
 
@@ -139,14 +147,11 @@ NATTEN is released under the [MIT License](LICENSE).
 
 ## Citation
 ```bibtex
-@article{hassani2022neighborhood,
+@inproceedings{hassani2023neighborhood,
 	title        = {Neighborhood Attention Transformer},
 	author       = {Ali Hassani and Steven Walton and Jiachen Li and Shen Li and Humphrey Shi},
-	year         = 2022,
-	url          = {https://arxiv.org/abs/2204.07143},
-	eprint       = {2204.07143},
-	archiveprefix = {arXiv},
-	primaryclass = {cs.CV}
+	year         = 2023,
+        booktitle    = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}
 }
 @article{hassani2022dilated,
 	title        = {Dilated Neighborhood Attention Transformer},

assets/README_pypi.md

@@ -50,18 +50,26 @@ We also welcome contributions in all forms.
 Just refer to our website, [shi-labs.com/natten](https://www.shi-labs.com/natten/), select your PyTorch version and the CUDA
 version it was compiled with, copy-paste the command and install in seconds!
 
-For example, if you're on `torch==1.12.1+cu116`, you should install NATTEN using the following wheel:
+For example, if you're on `torch==2.0.0+cu118`, you should install NATTEN using the following wheel:
 ```bash
-pip3 install natten -f https://shi-labs.com/natten/wheels/cu116/torch1.12.1/index.html
+pip3 install natten -f https://shi-labs.com/natten/wheels/cu118/torch2.0.0/index.html
 ```
 
 More generally:
 ```bash
 pip3 install natten -f https://shi-labs.com/natten/wheels/{cu_version}/torch{torch_version}/index.html
 ```
 
-**NOTE:** If you do not specify a wheel URL, you will install a "placeholder" version of NATTEN, which is not usable.
-We strongly recommend using our website, or building from source.
+**NOTE:** If you do not specify a wheel URL, pip will collect NATTEN and try to compile on locally, which depending
+on your system might take up to 30 minutes.
+We strongly recommend using our website if you're a Linux user.
+
+### Mac
+Unfortunately we are not yet able to build Mac wheels, but you can compile on install, so just run:
+
+```bash
+pip3 install natten
+```
 
 ### Windows
 NATTEN should support Windows devices with CUDA, but does not yet have Windows wheels.
@@ -74,7 +82,13 @@ Once you've set up your Python environment and installed PyTorch with CUDA, simp
 pip install ninja # Recommended, not required
 git clone https://github.com/SHI-Labs/NATTEN
 cd NATTEN
-pip install -e .
+make
+```
+
+#### Optional: run unit tests
+You can optionally run unit tests to verify building from source finished successfully:
+```bash
+make test
 ```
 
 
@@ -133,14 +147,11 @@ NATTEN is released under the [MIT License](https://github.com/SHI-Labs/NATTEN/bl
 
 ## Citation
 ```bibtex
-@article{hassani2022neighborhood,
+@inproceedings{hassani2023neighborhood,
 	title        = {Neighborhood Attention Transformer},
 	author       = {Ali Hassani and Steven Walton and Jiachen Li and Shen Li and Humphrey Shi},
-	year         = 2022,
-	url          = {https://arxiv.org/abs/2204.07143},
-	eprint       = {2204.07143},
-	archiveprefix = {arXiv},
-	primaryclass = {cs.CV}
+	year         = 2023,
+        booktitle    = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}
 }
 @article{hassani2022dilated,
 	title        = {Dilated Neighborhood Attention Transformer},

requirements.txt

@@ -1,7 +1,7 @@
--f https://download.pytorch.org/whl/cu117
-torch==1.13.0
--f https://download.pytorch.org/whl/cu117
-torchvision==0.14.0
+-f https://download.pytorch.org/whl/cu118
+torch==2.0.0
+-f https://download.pytorch.org/whl/cu118
+torchvision==0.15.1
 ninja
 flake8
 black

src/natten/__init__.py

@@ -1,10 +1,27 @@
-"""
-Neighborhood Attention
+#################################################################################################
+# Copyright (c) 2023 Ali Hassani.
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all
+# copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE.
+#
+#################################################################################################
 
-This source code is licensed under the license found in the
-LICENSE file in the root directory of this source tree.
-"""
 from .natten1d import NeighborhoodAttention1D
 from .natten2d import NeighborhoodAttention2D
 
-__version__ = "0.14.5"
+__version__ = "0.14.6"

src/natten/csrc/context.h

@@ -0,0 +1,27 @@
+/***************************************************************************************************
+ * Copyright (c) 2023 Ali Hassani.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ **************************************************************************************************/
+
+namespace natten {
+
+} // namespace natten
+

src/natten/csrc/cpu/natten1dav_cpu_kernel.cpp

@@ -1,17 +1,36 @@
-/*
-NATTEN1D-AV TORCH EXTENSION (CPU)
+/***************************************************************************************************
+ * Copyright (c) 2023 Ali Hassani.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ **************************************************************************************************/
 
-This source code is licensed under the license found in the
-LICENSE file in the root directory of this source tree.
-*/
 #include <torch/extension.h>
 #include <vector>
 #include <ATen/ATen.h>
 #include <ATen/AccumulateType.h>
+
 #if defined(AVX_INT)
 #include <ATen/cpu/vec/functional.h>
 #include <ATen/cpu/vec/vec.h>
 #endif
+
 #include "natten_cpu_commons.h"
 
 namespace natten {

src/natten/csrc/cpu/natten1dqkrpb_cpu_kernel.cpp

@@ -1,17 +1,36 @@
-/*
-NATTEN1D-QKRPB TORCH EXTENSION (CPU)
+/***************************************************************************************************
+ * Copyright (c) 2023 Ali Hassani.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ **************************************************************************************************/
 
-This source code is licensed under the license found in the
-LICENSE file in the root directory of this source tree.
-*/
 #include <torch/extension.h>
 #include <vector>
 #include <ATen/ATen.h>
 #include <ATen/AccumulateType.h>
+
 #if defined(AVX_INT)
 #include <ATen/cpu/vec/functional.h>
 #include <ATen/cpu/vec/vec.h>
 #endif
+
 #include "natten_cpu_commons.h"
 
 namespace natten {

src/natten/csrc/cpu/natten2dav_cpu_kernel.cpp

@@ -1,17 +1,36 @@
-/*
-NATTEN2D-AV TORCH EXTENSION (CPU)
+/***************************************************************************************************
+ * Copyright (c) 2023 Ali Hassani.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ **************************************************************************************************/
 
-This source code is licensed under the license found in the
-LICENSE file in the root directory of this source tree.
-*/
 #include <torch/extension.h>
 #include <vector>
 #include <ATen/ATen.h>
 #include <ATen/AccumulateType.h>
+
 #if defined(AVX_INT)
 #include <ATen/cpu/vec/functional.h>
 #include <ATen/cpu/vec/vec.h>
 #endif
+
 #include "natten_cpu_commons.h"
 
 namespace natten {