youngkingdom/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Updates to Python interface for PyPI packaging (#1209)

Browse Source 
* Updates

* Updates to notebooks
This commit is contained in:
Jack Kosaian
2023-11-28 12:52:12 -06:00
committed by GitHub
parent b5d8a5d9cc
commit 8098336d51
11 changed files with 279 additions and 59 deletions
Download Patch File Download Diff File Expand all files Collapse all files

44
examples/python/00_basic_gemm.ipynb
View File

@ -7,7 +7,46 @@
"metadata": {},
"source": [
"# Basic example of using the CUTLASS Python interface\n",
"This notebook walks through a basic example of using the CUTLASS Python interface to declare, compile, and run GEMMs.\n"
"This notebook walks through a basic example of using the CUTLASS Python interface to declare, compile, and run GEMMs.\n",
"\n",
"[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/NVIDIA/cutlass/tree/master/examples/python/00_basic_gemm.ipynb)\n"
]
},
{
"cell_type": "markdown",
"id": "df94d7e6",
"metadata": {},
"source": [
"## Prerequisites for running on Colab\n",
"This notebook requires an NVIDIA GPU. If `nvidia-smi` fails, go to Runtime -> Change runtime type -> Hardware accelerator and confirm a GPU is selected."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "71c7a069",
"metadata": {},
"outputs": [],
"source": [
"!#nvidia-smi"
]
},
{
"cell_type": "markdown",
"id": "cf16785d",
"metadata": {},
"source": [
"If running on Colab, you will need to install the CUTLASS Python interface. To do so, uncomment the following line and run the cell:"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "c819bb68",
"metadata": {},
"outputs": [],
"source": [
"!#pip install nvidia-cutlass"
]
},
{
@ -16,7 +55,8 @@
"id": "962324fd",
"metadata": {},
"source": [
"We first import various packages needed for the example and construct the input and output tensors that will be used in our example.\n"
"## General setup\n",
"We first import various packages needed for the example and construct the input and output tensors that will be used in our example."
]
},
{

45
examples/python/01_epilogue.ipynb
View File

@ -7,14 +7,55 @@
"metadata": {},
"source": [
"# Example of using elementwise activation functions in the CUTLASS Python interface\n",
"This notebook walks through a basic example of using the CUTLASS Python interface to declare, compile, and run GEMMs with different epilogues.\n"
"This notebook walks through a basic example of using the CUTLASS Python interface to declare, compile, and run GEMMs with different epilogues.\n",
"\n",
"[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/NVIDIA/cutlass/tree/master/examples/python/01_epilogue.ipynb)\n"
]
},
{
"cell_type": "markdown",
"id": "3ca993fe",
"id": "28c916da",
"metadata": {},
"source": [
"## Prerequisites for running on Colab\n",
"This notebook requires an NVIDIA GPU. If `nvidia-smi` fails, go to Runtime -> Change runtime type -> Hardware accelerator and confirm a GPU is selected."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "0fcea8ea",
"metadata": {},
"outputs": [],
"source": [
"!#nvidia-smi"
]
},
{
"cell_type": "markdown",
"id": "7ec60b57",
"metadata": {},
"source": [
"If running on Colab, you will need to install the CUTLASS Python interface. To do so, uncomment the following line and run the cell:"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "1db9e51c",
"metadata": {},
"outputs": [],
"source": [
"!#pip install nvidia-cutlass"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "962324fd",
"metadata": {},
"source": [
"## General setup\n",
"We first import various packages needed for the example and construct the input and output tensors that will be used in our example."
]
},

46
examples/python/02_pytorch_extension_grouped_gemm.ipynb
View File

@ -10,6 +10,52 @@
"This notebook walks through a basic example of using the CUTLASS Python interface to declare\n",
"a grouped GEMM kernel and export it as a PyTorch CUDA extension. Note that GEMM and Conv2d can also be exported as PyTorch CUDA extensions. \n",
"\n",
"[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/NVIDIA/cutlass/tree/master/examples/python/02_pytorch_extension_grouped_gemm.ipynb)\n"
]
},
{
"cell_type": "markdown",
"id": "2d70560e",
"metadata": {},
"source": [
"## Prerequisites for running on Colab\n",
"This notebook requires an NVIDIA GPU. If `nvidia-smi` fails, go to Runtime -> Change runtime type -> Hardware accelerator and confirm a GPU is selected."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "cc7c7458",
"metadata": {},
"outputs": [],
"source": [
"!#nvidia-smi"
]
},
{
"cell_type": "markdown",
"id": "2107bb0d",
"metadata": {},
"source": [
"If running on Colab, you will need to install the CUTLASS Python interface and PyTorch. To do so, uncomment the following line and run the cell:"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "a9852cb8",
"metadata": {},
"outputs": [],
"source": [
"!#pip install nvidia-cutlass torch --extra-index-url https://download.pytorch.org/whl/cu121"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "962324fd",
"metadata": {},
"source": [
"## Background on grouped GEMM\n",
"Grouped GEMM enables one to execute a set of GEMMs (each with potentially different sizes and strides)\n",
"in a single CUDA kernel. It can be thought of as a generalized version of a pointer-array GEMM,\n",

42
examples/python/03_basic_conv2d.ipynb
View File

@ -8,6 +8,48 @@
"\n",
"This notebook walks through a basic example of using the CUTLASS Python interface to declare, compile, and run Conv2d. \n",
"\n",
"[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/NVIDIA/cutlass/tree/master/examples/python/03_basic_conv2d.ipynb)\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Prerequisites for running on Colab\n",
"This notebook requires an NVIDIA GPU. If `nvidia-smi` fails, go to Runtime -> Change runtime type -> Hardware accelerator and confirm a GPU is selected."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"!#nvidia-smi"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"If running on Colab, you will need to install the CUTLASS Python interface. To do so, uncomment the following line and run the cell:"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"!#pip install nvidia-cutlass"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"## General setup\n",
"We first import various packages needed for the example and construct the input and output tensors that will be used in our example."
]
},

45
examples/python/04_epilogue_visitor.ipynb
View File

@ -7,14 +7,55 @@
"metadata": {},
"source": [
"# Example of using epilogue visitor in the CUTLASS Python interface\n",
"This notebook walks through a basic example of using the CUTLASS Python interface to declare, compile, and run GEMMs with different epilogues through CUTLASS Epilogue Visitor."
"This notebook walks through a basic example of using the CUTLASS Python interface to declare, compile, and run GEMMs with different epilogues through CUTLASS Epilogue Visitor.\n",
"\n",
"[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/NVIDIA/cutlass/tree/master/examples/python/04_epilogue_visitor.ipynb)\n"
]
},
{
"cell_type": "markdown",
"id": "3ca993fe",
"id": "3a800e79",
"metadata": {},
"source": [
"## Prerequisites for running on Colab\n",
"This notebook requires an NVIDIA GPU. If `nvidia-smi` fails, go to Runtime -> Change runtime type -> Hardware accelerator and confirm a GPU is selected."
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "9cfff2c8",
"metadata": {},
"outputs": [],
"source": [
"!#nvidia-smi"
]
},
{
"cell_type": "markdown",
"id": "06706f00",
"metadata": {},
"source": [
"If running on Colab, you will need to install the CUTLASS Python interface. To do so, uncomment the following line and run the cell:"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "491a7314",
"metadata": {},
"outputs": [],
"source": [
"!#pip install nvidia-cutlass"
]
},
{
"attachments": {},
"cell_type": "markdown",
"id": "962324fd",
"metadata": {},
"source": [
"## General setup\n",
"We first import various packages needed for the example, construct the input and output tensors that will be used in our example."
]
},