NVIDIA/apex issues and pull requests

#1874 - [Orin Nano] Error when trying to use sparsity

Issue - State: open - Opened by kyflores 10 days ago
Labels: bug

#1873 - multi_tensor_l2norm_scale_kernel compilation error

Issue - State: closed - Opened by zslefour 12 days ago

#1872 - multi_tensor_l2norm_scale_kernel compilation error

Issue - State: closed - Opened by zslefour 12 days ago

#1871 - as: symbol lookup error: as: undefined symbol: ZSTD_compressStream2

Issue - State: open - Opened by KLL535 14 days ago

#1870 - import apex always have AttributeError: module 'torch' has no attribute 'library'

Issue - State: open - Opened by Cominder 14 days ago - 7 comments

#1869 - Fix the typo of FusedRMSNorm doc

Pull Request - State: open - Opened by cqulilujia 19 days ago

#1868 - How should I search for the corresponding APEX release for a specific PyTorch version?

Issue - State: open - Opened by skyrise-l 19 days ago - 3 comments

#1867 - [FutureWarning] Resolving warnings caused by deprecated function usage.

Pull Request - State: open - Opened by drivanov about 1 month ago

#1866 - bug: FastLayerNorm

Issue - State: closed - Opened by KK666-AI about 1 month ago - 2 comments
Labels: bug

#1865 - Build sdist and wheel in CI

Pull Request - State: open - Opened by calebho about 1 month ago - 1 comment

#1864 - Traceable LayerNorm

Pull Request - State: closed - Opened by yanboliang about 1 month ago

#1863 - Use Mem Pool API for NCCL Zero-Copy

Pull Request - State: closed - Opened by Aidyn-A about 1 month ago - 1 comment

#1862 - Traceable RMSNorm part 2: non affine fused rms norm

Pull Request - State: closed - Opened by yanboliang about 2 months ago - 7 comments

#1861 - Traceable RMSNorm

Pull Request - State: closed - Opened by yanboliang about 2 months ago

#1860 - [NCCL] Updated fix for premature destruction of ProcessGroupNCCL

Pull Request - State: closed - Opened by eqy about 2 months ago

#1859 - [NCCL] Prevent premature destroy of PGs following PyTorch upstream change

Pull Request - State: closed - Opened by eqy about 2 months ago

#1858 - [GroupNorm] Skip GroupNorm tests on A16, A2 etc.,

Pull Request - State: closed - Opened by eqy about 2 months ago

#1857 - AttributeError: module 'torch.compiler' has no attribute 'is_compiling'

Issue - State: open - Opened by LukeLIN-web 2 months ago - 5 comments
Labels: bug

#1856 - Failed the last time, succeeded the next time?上一次还失败，下一次就成功了？

Issue - State: open - Opened by zhangs-a-n 3 months ago - 3 comments

#1855 - `Tensor.type()` -> `Tensor.scalar_type()`

Pull Request - State: closed - Opened by crcrpar 3 months ago

#1854 - [PT2] Normalisation: use manual impl when compiling

Pull Request - State: closed - Opened by alexdremov 3 months ago - 1 comment

#1853 - FusedRMSNormAffineMixedDtypesFunction is not importable in the PyTorch build without distributed support

Issue - State: open - Opened by IvanYashchuk 3 months ago
Labels: bug

#1852 - 关于解决ModuleNotFoundError: No module named 'torch'导致安装失败

Issue - State: open - Opened by Eikwang 3 months ago - 26 comments
Labels: bug

#1851 - How to install apex

Issue - State: open - Opened by Zerycii 3 months ago - 12 comments

#1850 - 到底要怎样才能安装apex

Issue - State: open - Opened by poyingshihuang 3 months ago - 1 comment

#1849 - AdamW implementation does not truly decouple learning rate and weight decay

Issue - State: open - Opened by leenachennuru 3 months ago - 2 comments
Labels: bug

#1848 - Failed to build installable wheels for some pyproject.toml based projects (apex)

Issue - State: open - Opened by RukaiaAfsana 4 months ago - 5 comments

#1847 - Avoid unnecessary NCCL collective coalescing in distributed optimizer

Pull Request - State: closed - Opened by timmoon10 4 months ago

#1846 - Literature associated with fused_dense

Issue - State: open - Opened by prmudgal 4 months ago - 1 comment

#1845 - fix groupnorm int32 index overflow

Pull Request - State: open - Opened by tlogn 4 months ago - 2 comments

#1844 - Main

Pull Request - State: closed - Opened by 63days 4 months ago

#1843 - ASP Automatic Sparsity forward function For Loop Error

Issue - State: open - Opened by maro-jeon 4 months ago

#1842 - Discrepancy with Optimizer States and Model State Dict when using store_param_remainders==True

Issue - State: open - Opened by alxzhang-amazon 4 months ago - 8 comments

#1841 - Gradient Overflow with Specific GPU Combinations in Multi-GPU Setup (NVIDIA RTX 3090)

Issue - State: open - Opened by SylU0 5 months ago

#1840 - No module named 'amp_C'

Issue - State: open - Opened by KanyuBao 5 months ago
Labels: bug

#1839 - loss scale

Issue - State: open - Opened by yjy-10 5 months ago

#1838 - How to improve training performance with Apex package

Issue - State: open - Opened by tjk9501 5 months ago

#1837 - Reformat Grad Output If It's Not Channels Last

Pull Request - State: closed - Opened by alpha0422 5 months ago
Labels: contrib

#1836 - Add Unittest For Distributed Adam With CUDA Graph

Pull Request - State: closed - Opened by alpha0422 5 months ago

#1835 - Traceable GroupNorm

Pull Request - State: closed - Opened by alpha0422 5 months ago
Labels: contrib

#1834 - install bug with pytorch2.0.1

Issue - State: open - Opened by Duanjinyi1 5 months ago - 1 comment
Labels: bug

#1833 - Unsupported NVHPC compiler found. nvc++ is the only NVHPC compiler that is supported.

Issue - State: closed - Opened by mz687 5 months ago
Labels: bug

#1832 - Enhance Distributed Fused Adam

Pull Request - State: closed - Opened by alpha0422 5 months ago
Labels: contrib

#1831 - Installation with Cuda extentions is failling

Issue - State: open - Opened by SaiedaJN 5 months ago - 2 comments
Labels: bug

#1830 - remove `run_transformer` from default lists

Pull Request - State: closed - Opened by crcrpar 5 months ago

#1829 - Fix DistributedTestBase for transformer distributed tests

Pull Request - State: closed - Opened by xwang233 5 months ago - 1 comment

#1828 - No CUDA runtime is found, using CUDA_HOME='/home/shengjieyi/cuda1108' .

Issue - State: open - Opened by vvsherryvv 5 months ago
Labels: bug

#1827 - Unsuccessful installation of apex library. (Preparing metadata (pyproject.toml) did not run successfully.)

Issue - State: open - Opened by ssaral 5 months ago - 4 comments
Labels: bug

#1826 - Slow Performance with "Exhaustive Search" Permutation Strategy for Channel Pruning in CNN

Issue - State: open - Opened by Ulorewien 5 months ago
Labels: bug

#1825 - Fix illegal memory access with multi_tensor_apply size above INT_MAX

Pull Request - State: closed - Opened by gdb 5 months ago - 3 comments

#1824 - Unable to install Apex

Issue - State: open - Opened by JoongunPark 5 months ago - 1 comment
Labels: bug

#1823 - Setting up Apex and get this error: ModuleNotFoundError: No module named 'torch'

Issue - State: closed - Opened by Mayolov 6 months ago - 12 comments
Labels: bug

#1822 - Install set.up

Issue - State: open - Opened by Maritime-Moon 6 months ago
Labels: bug

#1821 - Allow Configurable Cache Directory

Pull Request - State: closed - Opened by leimao 6 months ago

#1820 - [Distributed optimizer] Do not monkey-patch class methods

Pull Request - State: closed - Opened by timmoon10 6 months ago

#1819 - Unknown CUDA arch (compute) or GPU not supported error while installing on docker ubuntu with cuda 12.1

Issue - State: open - Opened by AvisP 6 months ago - 1 comment

#1818 - NCCLAllocator: Fix build failure

Pull Request - State: closed - Opened by Aidyn-A 6 months ago - 4 comments

#1817 - Unable to install Apex on Linux(debian) with CUDA 12.1 and torch 2.2.2

Issue - State: open - Opened by SamitM1 6 months ago - 2 comments
Labels: bug

#1816 - Release GIL

Pull Request - State: closed - Opened by crcrpar 7 months ago - 1 comment

#1815 - unable to install

Issue - State: open - Opened by lxy51 7 months ago - 5 comments

#1814 - Release GIL when calling C extensions

Issue - State: closed - Opened by szmigacz 7 months ago
Labels: bug

#1813 - deprecate uses of torch.cuda.amp

Pull Request - State: closed - Opened by Fuzzkatt 7 months ago - 2 comments

#1812 - Only print the warning message about `TORCH_CUDA_ARCH_LIST` if not set

Pull Request - State: open - Opened by aurianer 7 months ago

#1811 - fixup concats for grouped convolution

Pull Request - State: open - Opened by techshoww 7 months ago

#1810 - Unable to install Apex

Issue - State: open - Opened by Anupam-5 7 months ago - 2 comments
Labels: bug

#1809 - Win11+Visual Studio 2022,install successfully.

Issue - State: open - Opened by aswordok 7 months ago - 3 comments
Labels: bug

#1808 - Fixed compute type for FP16 Tensor core wrapper around cublas GEMMEx

Pull Request - State: closed - Opened by suachong 7 months ago

#1807 - Error

Issue - State: open - Opened by silentghost1412 7 months ago

#1806 - Use torch.testing.all_close instead of get_max_diff in test_lamb.py

Pull Request - State: closed - Opened by Fuzzkatt 7 months ago - 1 comment

#1805 - "packaging" library exists but not found

Issue - State: closed - Opened by mahmoodn 7 months ago - 3 comments

#1804 - Cannot import name 'UnencryptedCookieSessionFactoryConfig'

Issue - State: closed - Opened by mahmoodn 7 months ago - 3 comments

#1803 - ImportError: cannot import name '_library_root_logger' from 'apex' (unknown location)

Issue - State: open - Opened by BBALU1660 8 months ago - 2 comments

#1802 - Contrib unit test failure in `openfold_triton/test_fused_adam_swa.py::FusedAdamSWATestCase::test_fused_update_on_random_data`

Issue - State: open - Opened by xwang233 8 months ago
Labels: bug

#1801 - Avoid importing apex transformer automatically

Pull Request - State: open - Opened by nWEIdia 8 months ago

#1800 - Not Able to install apex.

Issue - State: open - Opened by Avinash-py 9 months ago
Labels: bug

#1799 - [ INSTALLATION ] - Not able to install apex on a Linux machine

Issue - State: open - Opened by MBadriNarayanan 9 months ago - 4 comments
Labels: bug

#1798 - Fix reduce_blocks_into_lanes race condition

Pull Request - State: closed - Opened by Fuzzkatt 9 months ago

#1797 - NCCL userbuffer for DP RS in DistOpt

Pull Request - State: closed - Opened by WanZzzzzz 9 months ago

#1796 - Add nccl_allocator for zero-copy user buffer

Pull Request - State: closed - Opened by Aidyn-A 9 months ago

#1795 - Avoid unnecessary param write in distributed Adam kernel

Pull Request - State: closed - Opened by timmoon10 9 months ago - 1 comment

#1794 - Enhance Distributed Fused Adam

Pull Request - State: closed - Opened by alpha0422 9 months ago - 4 comments

#1793 - apex not installing

Issue - State: open - Opened by pradeepdev-1995 10 months ago
Labels: bug

#1792 - Up to date patch for Windows compilation with Visual Studio 2022, CUDA 12.1 and PyTorch 2.2.2

Issue - State: open - Opened by doctorpangloss 10 months ago - 3 comments
Labels: bug

#1791 - fix building torch extension with glog

Pull Request - State: open - Opened by petronny 10 months ago - 1 comment

#1790 - Add xentropy bf16 support

Pull Request - State: open - Opened by zyeric 10 months ago

#1789 - Unclear licensing for contrib/sparsity

Issue - State: open - Opened by hyandell 10 months ago
Labels: bug

#1788 - install failure

Issue - State: open - Opened by 52Hzaaa 10 months ago
Labels: bug

#1787 - No module named 'torch._six'

Issue - State: open - Opened by xujin1184104394 10 months ago - 1 comment
Labels: bug

#1786 - 64-bit indexing Adam

Pull Request - State: open - Opened by cdm114514 10 months ago

#1785 - cannot import name 'AutoencoderKLTemporalDecoder' from 'diffusers.models'

Issue - State: closed - Opened by zj19941113 10 months ago
Labels: bug

#1784 - Add 2D Fused RoPE

Pull Request - State: closed - Opened by yaox12 10 months ago

#1783 - Move to the correct device for v1 state dict

Pull Request - State: closed - Opened by acphile 10 months ago - 2 comments
Labels: contrib

#1782 - Bump thresholds for `test_backward` in `test_fused_softmax.py`

Pull Request - State: closed - Opened by eqy 10 months ago

#1781 - On installing apex (+without sudo/docker)

Issue - State: open - Opened by stet-stet 11 months ago - 1 comment
Labels: bug

#1780 - [Questing] For apex sparsity model, when i export trt engine with flag sparsity=enable or force, only partial layer picked sparse implementation.

Issue - State: closed - Opened by Bobo-y 11 months ago - 4 comments

#1779 - Installation Problem: RuntimeError: Error compiling objects for extension

Issue - State: open - Opened by wwma 11 months ago - 3 comments
Labels: bug

#1778 - Cannot compile/build cuda_ext on H100

Issue - State: open - Opened by GuanhuaWang 11 months ago
Labels: bug

#1777 - [CUDNN][cudnn-frontend] Bump cuDNN to 1.0.3

Pull Request - State: open - Opened by eqy 11 months ago

#1776 - memory format option is only supported by strided tensors

Issue - State: closed - Opened by Cheny1m 12 months ago - 1 comment
Labels: bug

#1775 - Skip the p2p test on single GPU platforms

Pull Request - State: closed - Opened by nWEIdia 12 months ago

GitHub / NVIDIA/apex issues and pull requests