Remove trailing whitespace & ensure newline at EOF

CHANGELOG.md

@@ -49,7 +49,7 @@ Features:
 Bug fixes:
  - Fixed a bug where weight decay was incorrectly applied to 32-bit Adam. #13
  - Fixed an unsafe use of eval. #8
- - Fixed a bug where the StableEmbedding layer 32-bit optimizer override would not work without registering the whole model first (`bnb.optim.GlobalOptimManager.get_instance().register_parameters(model.parameters())`).  #13 #15 
+ - Fixed a bug where the StableEmbedding layer 32-bit optimizer override would not work without registering the whole model first (`bnb.optim.GlobalOptimManager.get_instance().register_parameters(model.parameters())`).  #13 #15
 
 Docs:
  - Added instructions how to solve "\_\_fatbinwrap_" errors.

CONTRIBUTING.md

@@ -28,4 +28,4 @@ outlined on that page and do not file a public issue.
 
 ## License
 By contributing to bitsandbytes, you agree that your contributions will be licensed
-under the LICENSE file in the root directory of this source tree.
+under the LICENSE file in the root directory of this source tree.

Makefile

@@ -26,14 +26,14 @@ INCLUDE :=  -I $(CUDA_HOME)/include -I $(ROOT_DIR)/csrc -I $(CONDA_PREFIX)/inclu
 LIB := -L $(CUDA_HOME)/lib64 -lcudart -lcublas -lcublasLt -lcurand -lcusparse -L $(CONDA_PREFIX)/lib
 
 # NVIDIA NVCC compilation flags
-COMPUTE_CAPABILITY := -gencode arch=compute_35,code=sm_35 # Kepler 
-COMPUTE_CAPABILITY += -gencode arch=compute_37,code=sm_37 # Kepler 
+COMPUTE_CAPABILITY := -gencode arch=compute_35,code=sm_35 # Kepler
+COMPUTE_CAPABILITY += -gencode arch=compute_37,code=sm_37 # Kepler
 COMPUTE_CAPABILITY += -gencode arch=compute_50,code=sm_50 # Maxwell
 COMPUTE_CAPABILITY += -gencode arch=compute_52,code=sm_52 # Maxwell
 COMPUTE_CAPABILITY += -gencode arch=compute_60,code=sm_60 # Pascal
 COMPUTE_CAPABILITY += -gencode arch=compute_61,code=sm_61 # Pascal
 COMPUTE_CAPABILITY += -gencode arch=compute_70,code=sm_70 # Volta
-COMPUTE_CAPABILITY += -gencode arch=compute_72,code=sm_72 # Volta 
+COMPUTE_CAPABILITY += -gencode arch=compute_72,code=sm_72 # Volta
 
 # CUDA 9.2 supports CC 3.0, but CUDA >= 11.0 does not
 CC_CUDA92 := -gencode arch=compute_30,code=sm_30
@@ -58,38 +58,38 @@ CC_cublasLt111 += -gencode arch=compute_86,code=sm_86
 
 
 all: $(ROOT_DIR)/dependencies/cub $(BUILD_DIR) env
-	$(NVCC) $(COMPUTE_CAPABILITY) -Xcompiler '-fPIC' --use_fast_math -Xptxas=-v -dc $(FILES_CUDA) $(INCLUDE) $(LIB) --output-directory $(BUILD_DIR) 
-	$(NVCC) $(COMPUTE_CAPABILITY) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o 
+	$(NVCC) $(COMPUTE_CAPABILITY) -Xcompiler '-fPIC' --use_fast_math -Xptxas=-v -dc $(FILES_CUDA) $(INCLUDE) $(LIB) --output-directory $(BUILD_DIR)
+	$(NVCC) $(COMPUTE_CAPABILITY) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o
 	$(GPP) -std=c++14 -DBUILD_CUDA -shared -fPIC $(INCLUDE) $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o $(BUILD_DIR)/link.o $(FILES_CPP) -o ./bitsandbytes/libbitsandbytes_cuda$(CUDA_VERSION).so $(LIB)
 
 cuda92: $(ROOT_DIR)/dependencies/cub $(BUILD_DIR) env
 	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA92) -Xcompiler '-fPIC' --use_fast_math -Xptxas=-v -dc $(FILES_CUDA) $(INCLUDE) $(LIB) --output-directory $(BUILD_DIR) -D NO_CUBLASLT
-	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA92) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o 
+	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA92) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o
 	$(GPP) -std=c++14 -DBUILD_CUDA -shared -fPIC $(INCLUDE) $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o $(BUILD_DIR)/link.o $(FILES_CPP) -o ./bitsandbytes/libbitsandbytes_cuda$(CUDA_VERSION)_nocublaslt.so $(LIB)
 
 cuda10x_nomatmul: $(ROOT_DIR)/dependencies/cub $(BUILD_DIR) env
 	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA10x) -Xcompiler '-fPIC' --use_fast_math -Xptxas=-v -dc $(FILES_CUDA) $(INCLUDE) $(LIB) --output-directory $(BUILD_DIR) -D NO_CUBLASLT
-	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA10x) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o 
+	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA10x) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o
 	$(GPP) -std=c++14 -DBUILD_CUDA -shared -fPIC $(INCLUDE) $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o $(BUILD_DIR)/link.o $(FILES_CPP) -o ./bitsandbytes/libbitsandbytes_cuda$(CUDA_VERSION)_nocublaslt.so $(LIB)
 
 cuda110_nomatmul: $(BUILD_DIR) env
 	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA110) -Xcompiler '-fPIC' --use_fast_math -Xptxas=-v -dc $(FILES_CUDA) $(INCLUDE) $(LIB) --output-directory $(BUILD_DIR) -D NO_CUBLASLT
-	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA110) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o 
+	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA110) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o
 	$(GPP) -std=c++14 -DBUILD_CUDA -shared -fPIC $(INCLUDE) $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o $(BUILD_DIR)/link.o $(FILES_CPP) -o ./bitsandbytes/libbitsandbytes_cuda$(CUDA_VERSION)_nocublaslt.so $(LIB)
 
 cuda11x_nomatmul: $(BUILD_DIR) env
 	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA11x) -Xcompiler '-fPIC' --use_fast_math -Xptxas=-v -dc $(FILES_CUDA) $(INCLUDE) $(LIB) --output-directory $(BUILD_DIR) -D NO_CUBLASLT
-	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA11x) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o 
+	$(NVCC) $(COMPUTE_CAPABILITY) $(CC_CUDA11x) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o
 	$(GPP) -std=c++14 -DBUILD_CUDA -shared -fPIC $(INCLUDE) $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o $(BUILD_DIR)/link.o $(FILES_CPP) -o ./bitsandbytes/libbitsandbytes_cuda$(CUDA_VERSION)_nocublaslt.so $(LIB)
 
 cuda110: $(BUILD_DIR) env
 	$(NVCC) $(CC_cublasLt110) -Xcompiler '-fPIC' --use_fast_math -Xptxas=-v -dc $(FILES_CUDA) $(INCLUDE) $(LIB) --output-directory $(BUILD_DIR)
-	$(NVCC) $(CC_cublasLt110) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o 
+	$(NVCC) $(CC_cublasLt110) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o
 	$(GPP) -std=c++14 -DBUILD_CUDA -shared -fPIC $(INCLUDE) $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o $(BUILD_DIR)/link.o $(FILES_CPP) -o ./bitsandbytes/libbitsandbytes_cuda$(CUDA_VERSION).so $(LIB)
 
 cuda11x: $(BUILD_DIR) env
 	$(NVCC) $(CC_cublasLt111) -Xcompiler '-fPIC' --use_fast_math -Xptxas=-v -dc $(FILES_CUDA) $(INCLUDE) $(LIB) --output-directory $(BUILD_DIR)
-	$(NVCC) $(CC_cublasLt111) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o 
+	$(NVCC) $(CC_cublasLt111) -Xcompiler '-fPIC' -dlink $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o -o $(BUILD_DIR)/link.o
 	$(GPP) -std=c++14 -DBUILD_CUDA -shared -fPIC $(INCLUDE) $(BUILD_DIR)/ops.o $(BUILD_DIR)/kernels.o $(BUILD_DIR)/link.o $(FILES_CPP) -o ./bitsandbytes/libbitsandbytes_cuda$(CUDA_VERSION).so $(LIB)
 
 cpuonly: $(BUILD_DIR) env
@@ -117,7 +117,7 @@ $(ROOT_DIR)/dependencies/cub:
 	cd dependencies/cub; git checkout 1.11.0
 
 clean:
-	rm build/* 
+	rm build/*
 
 cleaneggs:
 	rm -rf *.egg*

README.md

@@ -1,6 +1,6 @@
 # bitsandbytes
 
-The bitsandbytes is a lightweight wrapper around CUDA custom functions, in particular 8-bit optimizers, matrix multiplication (LLM.int8()), and quantization functions. 
+The bitsandbytes is a lightweight wrapper around CUDA custom functions, in particular 8-bit optimizers, matrix multiplication (LLM.int8()), and quantization functions.
 
 
 
@@ -48,7 +48,7 @@ out = linear(x.to(torch.float16))
 
 Requirements: anaconda, cudatoolkit, pytorch
 
-Hardware requirements: 
+Hardware requirements:
  - LLM.int8(): NVIDIA Turing (RTX 20xx; T4) or Ampere GPU (RTX 30xx; A4-A100); (a GPU from 2018 or older).
  - 8-bit optimizers and quantization: NVIDIA Maxwell GPU or newer (>=GTX 9XX).
 
@@ -87,7 +87,7 @@ Note that by default all parameter tensors with less than 4096 elements are kept
 ```
 # parameter tensors with less than 16384 values are optimized in 32-bit
 # it is recommended to use multiplies of 4096
-adam = bnb.optim.Adam8bit(model.parameters(), min_8bit_size=16384) 
+adam = bnb.optim.Adam8bit(model.parameters(), min_8bit_size=16384)
 ```
 
 ### Change Bits and other Hyperparameters for Individual Parameters

bitsandbytes/autograd/_functions.py

@@ -15,7 +15,7 @@ def prod(iterable):
 
 """
     This class pools outlier dimensions across layers.
-    This is particularly important for small models where outlier features 
+    This is particularly important for small models where outlier features
     are less systematic and occur with low frequency.
 """
 class GlobalOutlierPooler(object):

bitsandbytes/functional.py

@@ -1133,7 +1133,7 @@ def igemm(
     ptr = CUBLAS_Context.get_instance().get_context(A.device)
 
     # B^T @ A^T = C^T
-    # [km, nk -> mn] 
+    # [km, nk -> mn]
     is_on_gpu([B, A, out])
     lib.cigemm(ptr, ct.c_bool(transposed_B), ct.c_bool(transposed_A), ct.c_int32(m), ct.c_int32(n), ct.c_int32(k),
                get_ptr(B), get_ptr(A), get_ptr(out), ct.c_int32(lda), ct.c_int32(ldb), ct.c_int32(ldc))

bitsandbytes/nn/modules.py

@@ -267,7 +267,7 @@ def forward(self, x):
                 self.weight.data = self.state.CxB
             elif self.state.memory_efficient_backward and self.state.CxB is not None:
                 # For memory efficient backward, we convert 8-bit row major to turing/ampere format at each inference pass.
-                # Thus, we delete CxB from the state. 
+                # Thus, we delete CxB from the state.
                 del self.state.CxB
 
         return out

compile_from_source.md

@@ -4,7 +4,7 @@ Basic steps.
 1. `make [target]` where `[target]` is among `cuda92, cuda10x, cuda110, cuda11x, cpuonly`
 2. `CUDA_VERSION=XXX python setup.py install`
 
-To run these steps you will need to have the nvcc compiler installed that comes with a CUDA installation. If you use anaconda (recommended) then you can figure out which version of CUDA you are using with PyTorch via the command `conda list | grep cudatoolkit`. Then you can install the nvcc compiler by downloading and installing the same CUDA version from the [CUDA toolkit archive](https://developer.nvidia.com/cuda-toolkit-archive). 
+To run these steps you will need to have the nvcc compiler installed that comes with a CUDA installation. If you use anaconda (recommended) then you can figure out which version of CUDA you are using with PyTorch via the command `conda list | grep cudatoolkit`. Then you can install the nvcc compiler by downloading and installing the same CUDA version from the [CUDA toolkit archive](https://developer.nvidia.com/cuda-toolkit-archive).
 
 For your convenience, there is an installation script in the root directory that installs CUDA 11.1 locally and configures it automatically. After installing you should add the `bin` sub-directory to the `$PATH` variable to make the compiler visible to your system. To do this you can add this to your `.bashrc` by executing these commands:
 ```bash
@@ -13,7 +13,7 @@ echo "export PATH=$PATH:/usr/local/cuda/bin/" >> ~/.bashrc
 source ~/.bashrc
 ```
 
-By default, the Makefile will look at your `CUDA_HOME` environmental variable to find your CUDA version for compiling the library. If this path is not set it is inferred from the path of your `nvcc` compiler. 
+By default, the Makefile will look at your `CUDA_HOME` environmental variable to find your CUDA version for compiling the library. If this path is not set it is inferred from the path of your `nvcc` compiler.
 
 Either `nvcc` needs to be in path for the `CUDA_HOME` variable needs to be set to the CUDA directory root (e.g. `/usr/local/cuda`) in order for compilation to succeed
 

csrc/cpu_ops.cpp

@@ -62,7 +62,7 @@ void quantize_cpu(float *code, float *A, float *absmax, unsigned char *out, long
 
       for (int i = 0; i < valid_chunks; i++)
           int err = pthread_join(threads[i], NULL);
-      
+
       free(threads);
       for (int i = 0; i < valid_chunks; i++)
           free(args[i]);