套件：libtorch-cuda-2.6（2.6.0+dfsg-6） [contrib]

[ 原始碼： pytorch-cuda  ]

Tensors and Dynamic neural networks in Python (Shared Objects)

PyTorch is a Python package that provides two high-level features:

(1) Tensor computation (like NumPy) with strong GPU acceleration (2) Deep neural networks built on a tape-based autograd system

You can reuse your favorite Python packages such as NumPy, SciPy and Cython to extend PyTorch when needed.

This is the CUDA version of PyTorch (Shared Objects).

其他與 libtorch-cuda-2.6 有關的套件

• dep: libblas3

  Basic Linear Algebra Reference implementations, shared library

  或者 libblas.so.3

  本虛擬套件由這些套件填實： libblas3, libblis4-openmp, libblis4-pthread, libblis4-serial, libopenblas0-openmp, libopenblas0-pthread, libopenblas0-serial

• dep: libc6 (>= 2.38)

  GNU C 函式庫：共用函式庫
  同時作為一個虛擬套件由這些套件填實： libc6-udeb

• dep: libcpp-httplib0.18 (>= 0.18.7)

  C++ HTTP/HTTPS server and client library

• dep: libcpuinfo0 (>= 0.0~git20230113.6481e8b~) [除 ppc64el]

  CPU INFOrmation library (shared object)

• dep: libcublas12

  NVIDIA cuBLAS Library

• dep: libcublaslt12

  NVIDIA cuBLASLt Library

• dep: libcuda1 (>= 525)

  NVIDIA CUDA Driver Library

  或者 libcuda.so.1 (>= 525)

  套件暫時不可用

• dep: libcudart12

  NVIDIA CUDA Runtime Library

• dep: libcufft11

  NVIDIA cuFFT Library

• dep: libcurand10

  NVIDIA cuRAND Library

• dep: libcusolver11

  NVIDIA cuSOLVER Library

• dep: libcusparse12

  NVIDIA cuSPARSE Library

• dep: libdnnl3.6 (>= 3.6.0~) [除 ppc64el]

  oneAPI Deep Neural Network Library (oneDNN) (lib)

• dep: libfmt10 (>= 10.1.1+ds1)

  fast type-safe C++ formatting library -- library

• dep: libgcc-s1 (>= 4.2) [ppc64el]

  GCC 支援函式庫

  dep: libgcc-s1 (>= 4.3) [amd64]

  dep: libgcc-s1 (>= 4.5) [arm64]

• dep: libgloo-cuda-0 (>= 0.0~git20231202.5354032)

  Collective communications library (shared object)

• dep: libgomp1 (>= 4.9)

  GCC OpenMP (GOMP) support library

• dep: liblapack3

  Library of linear algebra routines 3 - shared version

  或者 liblapack.so.3

  本虛擬套件由這些套件填實： liblapack3, libopenblas0-openmp, libopenblas0-pthread, libopenblas0-serial

• dep: libnccl2 (>= 2.22.3)

  NVIDIA Optimized primitives for inter-GPU communication

• dep: libnuma1 (>= 2.0.11)

  Libraries for controlling NUMA policy

• dep: libnvrtc12

  CUDA Runtime Compilation (NVIDIA NVRTC Library)

• dep: libnvtoolsext1 (>= 5.0)

  NVIDIA Tools Extension Library

• dep: libonnx1t64 (>= 1.17.0)

  Open Neural Network Exchange (ONNX) (libs)

• dep: libprotobuf32t64 (>= 3.21.12)

  protocol buffers C++ library

• dep: libpthreadpool0 (>= 0.0~git20210507.1787867~) [除 ppc64el]

  pthread-based thread pool for C/C++ (shared object)

• dep: libsleef3 (>= 3.6.1) [除 amd64]

  SLEEF Vectorized Math Library (libraries)

  dep: libsleef3 (>= 3.6.1-1~) [amd64]

• dep: libstdc++6 (>= 12)

  GNU Standard C++ Library v3

• dep: libtensorpipe-cuda-0 (>= 0.0~git20220513.bb1473a)

  tensor-aware point-to-point communication primitive for machine learning

• dep: libvulkan1 (>= 1.2.131.2)

  Vulkan loader library

• dep: libxnnpack0.20241108 (>= 0.0~git20241108.4ea82e5) [除 ppc64el]

  High-efficiency floating-point neural network inference operators (libs)

• dep: libz3-4 (>= 4.13.3)

  theorem prover from Microsoft Research - runtime libraries

• dep: libzstd1 (>= 1.5.5)

  fast lossless compression algorithm

• dep: nvidia-cudnn (>= 9.0.0.312~)

  NVIDIA CUDA Deep Neural Network library (install script)

• dep: zlib1g (>= 1:1.2.0)

  壓縮函式庫 - 跑程式時用(runtime)

• rec: libopenblas0

  Optimized BLAS (linear algebra) library (meta)

  或者 libblis3

  套件暫時不可用

  或者 libatlas3-base

  套件暫時不可用

  或者 libmkl-rt

  Intel® MKL: Single Dynamic Library (SDL)

  或者 libblas3

  Basic Linear Algebra Reference implementations, shared library

下載 libtorch-cuda-2.6

下載可用於所有硬體架構的

硬體架構	套件大小	安裝後大小	檔案
amd64	482，388。9 kB	1，425，426。0 kB	[檔案列表]
arm64	467，551。2 kB	1，373，783。0 kB	[檔案列表]
ppc64el	474，378。4 kB	1，434，224。0 kB	[檔案列表]