Merge branch 'main' of github.com:facebookresearch/bitsandbytes into main

6 files changed, 15 insertions, 27 deletions

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 132e7ec..a5b29d8 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -21,3 +21,4 @@ Features:
 v0.0.24:
 
 - Fixed a bug where a float/half conversion led to a compilation error for CUDA 11.1 on Turning GPUs.
+- removed Apex dependency for bnb LAMB
diff --git a/Makefile b/Makefile
index 5f5efed..753ee3d 100644
--- a/Makefile
+++ b/Makefile
@@ -52,6 +52,7 @@ $(BUILD_DIR):
 
 $(ROOT_DIR)/dependencies/cub:
 	git clone https://github.com/NVlabs/cub $(ROOT_DIR)/dependencies/cub
+	cd dependencies/cub; git checkout 1.11.0
 
 clean:
 	rm cuda_build/* ./bitsandbytes/libbitsandbytes.so
diff --git a/bitsandbytes/optim/lamb.py b/bitsandbytes/optim/lamb.py
index b8d4b1e..58cc13d 100644
--- a/bitsandbytes/optim/lamb.py
+++ b/bitsandbytes/optim/lamb.py
@@ -2,7 +2,6 @@
 #   
 # This source code is licensed under the MIT license found in the 
 # LICENSE file in the root directory of this source tree.
-import apex
 from bitsandbytes.optim.optimizer import Optimizer2State
 
 class LAMB(Optimizer2State):
diff --git a/deploy.sh b/deploy.sh
deleted file mode 100644
index a08351e..0000000
--- a/deploy.sh
+++ /dev/null
@@ -1,13 +0,0 @@
-#!/bin/bash
-
-rm -rf dist build
-make clean
-CUDA_HOME=/usr/local/cuda-10.2 make
-CUDA_VERSION=102 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
-
-rm -rf dist build
-make clean
-CUDA_HOME=/usr/local/cuda-11.1 make
-CUDA_VERSION=111 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
diff --git a/deploy_from_slurm.sh b/deploy_from_slurm.sh
index e21f2e0..6357e1d 100644
--- a/deploy_from_slurm.sh
+++ b/deploy_from_slurm.sh
@@ -10,7 +10,7 @@ module load gcc/7.3.0
 CUDA_HOME=/public/apps/cuda/9.2
 make
 CUDA_VERSION=92 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
+python -m twine upload dist/* --verbose
 module unload cuda
 
 
@@ -21,7 +21,7 @@ module load cuda/10.0
 CUDA_HOME=/public/apps/cuda/10.0
 make cuda10x
 CUDA_VERSION=100 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
+python -m twine upload dist/* --verbose
 module unload cuda
 module unload gcc
 module load gcc/8.4
@@ -33,7 +33,7 @@ module load cuda/10.1
 CUDA_HOME=/public/apps/cuda/10.1
 make cuda10x
 CUDA_VERSION=101 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
+python -m twine upload dist/* --verbose
 module unload cuda
 
 rm -rf dist build
@@ -43,7 +43,7 @@ module load cuda/10.2
 CUDA_HOME=/public/apps/cuda/10.2/
 make cuda10x
 CUDA_VERSION=102 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
+python -m twine upload dist/* --verbose
 module unload cuda
 
 
@@ -54,7 +54,7 @@ module load cuda/11.0
 CUDA_HOME=/public/apps/cuda/11.0
 make cuda110
 CUDA_VERSION=110 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
+python -m twine upload dist/* --verbose
 module unload cuda
 
 rm -rf dist build
@@ -64,7 +64,7 @@ module load cuda/11.1
 CUDA_HOME=/public/apps/cuda/11.1
 make cuda11x
 CUDA_VERSION=111 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
+python -m twine upload dist/* --verbose
 module unload cuda
 
 rm -rf dist build
@@ -74,7 +74,7 @@ module load cuda/11.2
 CUDA_HOME=/public/apps/cuda/11.2
 make cuda11x
 CUDA_VERSION=112 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
+python -m twine upload dist/* --verbose
 module unload cuda
 
 rm -rf dist build
@@ -82,5 +82,5 @@ make clean
 make cleaneggs
 CUDA_HOME=/private/home/timdettmers/git/autoswap/local/cuda-11.3 make cuda11x
 CUDA_VERSION=113 python -m build
-python -m twine upload --repository testpypi dist/* --verbose
+python -m twine upload dist/* --verbose
 module unload cuda
diff --git a/setup.py b/setup.py
index dc1eb60..36d1072 100644
--- a/setup.py
+++ b/setup.py
@@ -13,19 +13,19 @@ def read(fname):
 
 setup(
     name = f"bitsandbytes-cuda{os.environ['CUDA_VERSION']}",
-    version = "0.0.23",
+    version = "0.0.24",
     author = "Tim Dettmers",
-    author_email = "tim.dettmers@gmail.com",
-    description = ("Numpy-like library for GPUs."),
+    author_email = "dettmers@cs.washington.edu",
+    description = ("8-bit optimizers and quantization routines."),
     license = "MIT",
-    keywords = "gpu",
+    keywords = "gpu optimizers optimization 8-bit quantization compression",
     url = "http://packages.python.org/bitsandbytes",
     packages=find_packages(),
     package_data={'': ['libbitsandbytes.so']},
     long_description=read('README.md'),
     long_description_content_type = 'text/markdown',
     classifiers=[
-        "Development Status :: 1 - Planning",
+        "Development Status :: 4 - Beta",
         'Topic :: Scientific/Engineering :: Artificial Intelligence'
     ],
 )