PyTorch | NVIDIA NGC

NVIDIA

PyTorch

Container

NVIDIA

PyTorch

PyTorch is a GPU accelerated tensor computational framework. Functionality can be extended with common Python libraries such as NumPy and SciPy. Automatic differentiation is done with a tape-based system at the functional and neural network layer levels.

NVIDIA AI Enterprise Supported

Layer	Label		Created
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	LABEL	`com.nvidia.build.ref=36e6c838d631e94a47406a4e952221247afcabc0`	12/06/2024 12:14 AM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ARG	`NVIDIA_BUILD_REF=36e6c838d631e94a47406a4e952221247afcabc0`	12/06/2024 12:14 AM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	LABEL	`com.nvidia.build.id=126674151`	12/06/2024 12:14 AM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ENV	`NVIDIA_BUILD_ID=126674151`	12/06/2024 12:14 AM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ARG	`NVIDIA_BUILD_ID=126674151`	12/06/2024 12:14 AM UTC
sha256:27e863284e2bbf0e6d582cc95f0eaa25c21203bb71d606beeb13c9e3f2acdbbf	COPY	`entrypoint.d/ /opt/nvidia/entrypoint.d/`	12/06/2024 12:14 AM UTC
sha256:e7b515f1bc1c9756c09463872682db7a08fbee9c41322f1acb2bed88a6700433	RUN	`NVIDIA_PYTORCH_VERSION=24.12 PYTORCH_BUILD_VERSION=2.6.0a0+df5bbc0 NVFUSER_BUILD_VERSION=0d33366 TARGETARCH=arm64 PYVER=3.12 PYVER_MAJMIN=312 L4T=1 ln -sf ${_CUDA_COMPAT_PATH}/lib.real ${_CUDA_COMPAT_PATH}/lib && echo ${_CUDA_COMPAT_PATH}/lib > /etc/ld.so.conf.d/00-cuda-compat.conf && ldconfig && rm -f ${_CUDA_COMPAT_PATH}/lib`	12/06/2024 12:14 AM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ENV	`CUDA_MODULE_LOADING=LAZY`	12/06/2024 12:14 AM UTC
sha256:a3ed95caeb02ffe68cdd9fd84406680ae93d633cb16422d00e8a7c22955b46d4	ENV	`TORCH_CUDNN_V8_API_ENABLED=1`	12/06/2024 12:14 AM UTC
sha256:4f4fb700ef54461cfa02571ae0db9a0dc1e0cdb5577484a6d75e68dc38e8acc1	RUN	`NVIDIA_PYTORCH_VERSION=24.12 PYTORCH_BUILD_VERSION=2.6.0a0+df5bbc0 NVFUSER_BUILD_VERSION=0d33366 TARGETARCH=arm64 PYVER=3.12 PYVER_MAJMIN=312 L4T=1 if [ "${L4T}" = "1" ]; then echo "Not installing Transformer Engine in iGPU container until Version variable is set"; else NVTE_BUILD_THREADS_PER_JOB=8 pip install --no-cache-dir --no-build-isolation git+https://github.com/NVIDIA/TransformerEngine.git@release_v${TRANSFORMER_ENGINE_VERSION}; fi`	12/06/2024 12:14 AM UTC