Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Initial Support Blackwell Build #1418

Open
wants to merge 8 commits into
base: main
Choose a base branch
from
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 5 additions & 3 deletions README.rst
Original file line number Diff line number Diff line change
Expand Up @@ -58,10 +58,12 @@ simplifying mixed precision training for users.
Highlights
==========

* Easy-to-use modules for building Transformer layers with FP8 support
* Easy-to-use modules for building Transformer layers with FP8 support and FP4 support
* Optimizations (e.g. fused kernels) for Transformer models
* Support for FP8 on NVIDIA Hopper and NVIDIA Ada GPUs
* Support for FP8 on NVIDIA Hopper, NVIDIA Ada GPUs and NVIDIA Blackwell GPUs
* Support for FP4 on NVIDIA Blackwell GPUs
* Support for optimizations across all precisions (FP16, BF16) on NVIDIA Ampere GPU architecture generations and later
* Support for (FP32) on NVIDIA Turing GPU architecture.

Examples
========
Expand Down Expand Up @@ -149,7 +151,7 @@ Installation
Pre-requisites
^^^^^^^^^^^^^^^^^^^^
* Linux x86_64
* CUDA 12.0+ for Hopper and CUDA 12.1+ for Ada
* CUDA 12.0+ for Hopper, CUDA 12.1+ for Ada and Cuda 12.7+ for Blackwell
* NVIDIA Driver supporting CUDA 12.0 or later
* cuDNN 8.1 or later
* For fused attention, CUDA 12.1 or later, NVIDIA Driver supporting CUDA 12.1 or later, and cuDNN 8.9 or later.
Expand Down
2 changes: 1 addition & 1 deletion build_tools/utils.py
Original file line number Diff line number Diff line change
Expand Up @@ -190,7 +190,7 @@ def cuda_path() -> Tuple[str, str]:

@functools.lru_cache(maxsize=None)
def cuda_archs() -> str:
return os.getenv("NVTE_CUDA_ARCHS", "70;80;89;90")
return os.getenv("NVTE_CUDA_ARCHS", "70;80;89;90;100;120")


def cuda_version() -> Tuple[int, ...]:
Expand Down
7 changes: 7 additions & 0 deletions examples/jax/encoder/common.py
Original file line number Diff line number Diff line change
Expand Up @@ -19,3 +19,10 @@ def is_fp8_supported():
"""Return if FP8 has hardware supported"""
gpu_arch = get_device_compute_capability(0)
return gpu_arch >= 90


@lru_cache
def is_fp4_supported():
"""Return if FP4 has hardware supported"""
gpu_arch = get_device_compute_capability(0)
return gpu_arch >= 100
2 changes: 1 addition & 1 deletion tests/cpp/CMakeLists.txt
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@
cmake_minimum_required(VERSION 3.18)

if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
set(CMAKE_CUDA_ARCHITECTURES 70 80 90)
set(CMAKE_CUDA_ARCHITECTURES 70 80 90 100 120)
endif()


Expand Down
2 changes: 1 addition & 1 deletion transformer_engine/common/CMakeLists.txt
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ cmake_minimum_required(VERSION 3.21)

# Language options
if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
set(CMAKE_CUDA_ARCHITECTURES 70 80 89 90)
set(CMAKE_CUDA_ARCHITECTURES 70 80 89 90 100 120)
endif()
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CUDA_STANDARD 17)
Expand Down