search

rapidsai / cuml

cuML - RAPIDS Machine Learning Library
https://docs.rapids.ai/api/cuml/stable/
Apache License 2.0
4.23k stars 532 forks source link

FAISS memory error with UMAP (WSL2) #4290

Open alexwilson1 opened 3 years ago

alexwilson1 commented 3 years ago

Describe the bug

I have a dataset of 800k items (768 dim vectors). UMAP will work with the full 800k dataset, and with smaller (randomly sampled) datasets of around 150k, but medium-sized datasets of size ~300k, 350k etc crash with this error.

Traceback (most recent call last):
  File "/opt/project/callbacks.py", line 744, in on_click
    umap_data_3D = umap_for_clustering.fit_transform(embedding_matrix)
  File "/opt/conda/envs/rapids/lib/python3.8/site-packages/cuml/internals/api_decorators.py", line 549, in inner_set_get
    ret_val = func(*args, **kwargs)
  File "cuml/manifold/umap.pyx", line 659, in cuml.manifold.umap.UMAP.fit_transform

  File "/opt/conda/envs/rapids/lib/python3.8/site-packages/cuml/internals/api_decorators.py", line 409, in inner_with_setters
    return func(*args, **kwargs)
  File "cuml/manifold/umap.pyx", line 600, in cuml.manifold.umap.UMAP.fit

RuntimeError: Error in virtual void faiss::gpu::StandardGpuResourcesImpl::initializeForDevice(int) at /home/conda/feedstock_root/build_artifacts/faiss-split_1618468126454/work/faiss/gpu/StandardGpuResources.cpp:285: Error: 'err == cudaSuccess' failed: failed to cudaHostAlloc 268435456 bytes for CPU <-> GPU async copy buffer (error 2 out of memory)

I'm using a Titan RTX GPU with 24GB memory and nvidia-smi is showing more than enough free memory for this operation before applying fit_transform:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.10       Driver Version: 510.10       CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA TITAN RTX   WDDM  | 00000000:01:00.0 Off |                  N/A |
| 41%   29C    P8    10W / 280W |   3678MiB / 24576MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A         4      C   Insufficient Permissions        N/A      |
+-----------------------------------------------------------------------------+

This is using parameters (n_components=3, n_neighbors=15, min_dist=0.0) to create the UMAP model and fit_transform operation to apply it.

Using rapidsai/rapidsai:21.10-cuda11.2-base-ubuntu18.04-py3.8 with torch==1.9.1+cu111 applied on top of the environment.

Any idea why this works for the large dataset and not intermediate sized datasets please?

viclafargue commented 3 years ago

Thank you for opening the issue. A minimal reproducer would be welcomed for a more informed answer. However, from available information, I see two possible causes. First, you may simply run out of memory (device or host memory). Or otherwise, the failure might happen because of an incompatibility between the CUDA version installed in the environment (11.6) and the one RAPIDS (and FAISS) was compiled with (11.2). You may be able to fix this by installing a 11.4 CUDA environment and RAPIDS software suite.

dantegd commented 3 years ago

@viclafargue the CUDA 11.6 version reported by SMI is just the maximum CUDA version supported by that driver, doesn't mean that the environment/locally CUDA 11.6 is being used

@alexwilson1 could you run https://github.com/rapidsai/cuml/blob/branch-21.12/print_env.sh and put the output in a reply here? That'll also help triage

alexwilson1 commented 3 years ago

@viclafargue the CUDA 11.6 version reported by SMI is just the maximum CUDA version supported by that driver, doesn't mean that the environment/locally CUDA 11.6 is being used

@alexwilson1 could you run https://github.com/rapidsai/cuml/blob/branch-21.12/print_env.sh and put the output in a reply here? That'll also help triage

Click here to see environment details

     **git***
     *REDACTED*
     **git submodules***

     ***OS Information***
     DISTRIB_ID=Ubuntu
     DISTRIB_RELEASE=18.04
     DISTRIB_CODENAME=bionic
     DISTRIB_DESCRIPTION="Ubuntu 18.04.6 LTS"
     NAME="Ubuntu"
     VERSION="18.04.6 LTS (Bionic Beaver)"
     ID=ubuntu
     ID_LIKE=debian
     PRETTY_NAME="Ubuntu 18.04.6 LTS"
     VERSION_ID="18.04"
     HOME_URL="https://www.ubuntu.com/"
     SUPPORT_URL="https://help.ubuntu.com/"
     BUG_REPORT_URL="https://bugs.launchpad.net/ubuntu/"
     PRIVACY_POLICY_URL="https://www.ubuntu.com/legal/terms-and-policies/privacy-policy"
     VERSION_CODENAME=bionic
     UBUNTU_CODENAME=bionic
     Linux 816b6aa61b81 5.10.60.1-microsoft-standard-WSL2 #1 SMP Wed Aug 25 23:20:18 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux

     ***GPU Information***
     Mon Oct 18 18:33:09 2021
     +-----------------------------------------------------------------------------+
     | NVIDIA-SMI 510.00       Driver Version: 510.10       CUDA Version: 11.6     |
     |-------------------------------+----------------------+----------------------+
     | GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
     | Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
     |                               |                      |               MIG M. |
     |===============================+======================+======================|
     |   0  NVIDIA TITAN RTX    On   | 00000000:01:00.0 Off |                  N/A |
     | 41%   29C    P8    10W / 280W |    316MiB / 24576MiB |     N/A      Default |
     |                               |                      |                  N/A |
     +-------------------------------+----------------------+----------------------+

     +-----------------------------------------------------------------------------+
     | Processes:                                                                  |
     |  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
     |        ID   ID                                                   Usage      |
     |=============================================================================|
     |  No running processes found                                                 |
     +-----------------------------------------------------------------------------+

     ***CPU***
     Architecture:        x86_64
     CPU op-mode(s):      32-bit, 64-bit
     Byte Order:          Little Endian
     CPU(s):              8
     On-line CPU(s) list: 0-7
     Thread(s) per core:  1
     Core(s) per socket:  8
     Socket(s):           1
     Vendor ID:           GenuineIntel
     CPU family:          6
     Model:               158
     Model name:          Intel(R) Core(TM) i7-9700K CPU @ 3.60GHz
     Stepping:            13
     CPU MHz:             3600.011
     BogoMIPS:            7200.02
     Virtualization:      VT-x
     Hypervisor vendor:   Microsoft
     Virtualization type: full
     L1d cache:           32K
     L1i cache:           32K
     L2 cache:            256K
     L3 cache:            12288K
     Flags:               fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology cpuid pni pclmulqdq vmx ssse3 fma cx16 pcid sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi ept vpid ept_ad fsgsbase bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt xsaveopt xsavec xgetbv1 xsaves md_clear flush_l1d arch_capabilities

     ***CMake***

     ***g++***
     /usr/bin/g++
     g++ (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0
     Copyright (C) 2017 Free Software Foundation, Inc.
     This is free software; see the source for copying conditions.  There is NO
     warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

     ***nvcc***

     ***Python***
     /opt/conda/envs/rapids/bin/python
     Python 3.8.12

     ***Environment Variables***
     PATH                            : /opt/conda/envs/rapids/bin:/opt/conda/condabin:/opt/conda/envs/rapids/bin:/opt/conda/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
     LD_LIBRARY_PATH                 : /usr/local/nvidia/lib:/usr/local/nvidia/lib64
     NUMBAPRO_NVVM                   :
     NUMBAPRO_LIBDEVICE              :
     CONDA_PREFIX                    : /opt/conda/envs/rapids
     PYTHON_PATH                     :

     ***conda packages***
     /opt/conda/condabin/conda
     # packages in environment at /opt/conda/envs/rapids:
     #
     # Name                    Version                   Build  Channel
     _libgcc_mutex             0.1                 conda_forge    conda-forge
     _openmp_mutex             4.5                      1_llvm    conda-forge
     abseil-cpp                20210324.2           h9c3ff4c_0    conda-forge
     aiohttp                   3.7.4.post0      py38h497a2fe_0    conda-forge
     anyio                     3.3.2            py38h578d9bd_0    conda-forge
     appdirs                   1.4.4              pyh9f0ad1d_0    conda-forge
     apscheduler               3.8.0            py38h578d9bd_0    conda-forge
     argon2-cffi               20.1.0           py38h497a2fe_2    conda-forge
     arrow-cpp                 5.0.0           py38h327e1ba_4_cuda    conda-forge
     arrow-cpp-proc            3.0.0                      cuda    conda-forge
     async-timeout             3.0.1                   py_1000    conda-forge
     async_generator           1.10                       py_0    conda-forge
     attrs                     21.2.0             pyhd8ed1ab_0    conda-forge
     aws-c-cal                 0.5.11               h95a6274_0    conda-forge
     aws-c-common              0.6.2                h7f98852_0    conda-forge
     aws-c-event-stream        0.2.7               h3541f99_13    conda-forge
     aws-c-io                  0.10.5               hfb6a706_0    conda-forge
     aws-checksums             0.1.11               ha31a3da_7    conda-forge
     aws-sdk-cpp               1.8.186              hb4091e7_3    conda-forge
     backcall                  0.2.0              pyh9f0ad1d_0    conda-forge
     backports                 1.0                        py_2    conda-forge
     backports.functools_lru_cache 1.6.4              pyhd8ed1ab_0    conda-forge
     blas                      2.108                       mkl    conda-forge
     blas-devel                3.9.0                     8_mkl    conda-forge
     blazingsql                21.10.0                  pypi_0    pypi
     bleach                    4.1.0              pyhd8ed1ab_0    conda-forge
     blosc                     1.21.0               h9c3ff4c_0    conda-forge
     bokeh                     2.4.0            py38h578d9bd_0    conda-forge
     boost                     1.72.0           py38h1e42940_1    conda-forge
     boost-cpp                 1.72.0               h312852a_5    conda-forge
     brotli                    1.0.9                h7f98852_5    conda-forge
     brotli-bin                1.0.9                h7f98852_5    conda-forge
     brotli-python             1.0.9            py38h709712a_5    conda-forge
     brotlipy                  0.7.0           py38h497a2fe_1001    conda-forge
     brunsli                   0.1                  h9c3ff4c_0    conda-forge
     bzip2                     1.0.8                h7f98852_4    conda-forge
     c-ares                    1.17.2               h7f98852_0    conda-forge
     ca-certificates           2021.10.8            ha878542_0    conda-forge
     cachetools                4.2.4              pyhd8ed1ab_0    conda-forge
     cairo                     1.16.0            h6cf1ce9_1008    conda-forge
     certifi                   2021.10.8        py38h578d9bd_0    conda-forge
     cffi                      1.14.6           py38h3931269_1    conda-forge
     cfitsio                   3.470                hb418390_7    conda-forge
     chardet                   4.0.0            py38h578d9bd_1    conda-forge
     charls                    2.2.0                h9c3ff4c_0    conda-forge
     charset-normalizer        2.0.0              pyhd8ed1ab_0    conda-forge
     chart-studio              1.1.0              pyh9f0ad1d_0    conda-forge
     click                     7.1.2              pyh9f0ad1d_0    conda-forge
     click-plugins             1.1.1                      py_0    conda-forge
     cligj                     0.7.2              pyhd8ed1ab_0    conda-forge
     cloudpickle               2.0.0              pyhd8ed1ab_0    conda-forge
     colorama                  0.4.4              pyh9f0ad1d_0    conda-forge
     colorcet                  2.0.6              pyhd8ed1ab_0    conda-forge
     cryptography              3.4.7            py38ha5dfef3_0    conda-forge
     cucim                     21.10.00        cuda_11.0_py38_gd7ac21f_0    rapidsai
     cudatoolkit               11.0.221             h6bb024c_0    nvidia
     cudf                      21.10.00        cuda_11.0_py38_g072fd862cc_0    rapidsai
     cudf_kafka                21.10.00        py38_g072fd862cc_0    rapidsai
     cugraph                   21.10.00        cuda11.0_py38_g84617024_0    rapidsai
     cuml                      21.10.00        cuda11.0_py38_g0fd3503ba_0    rapidsai
     cupy                      9.0.0            py38hc350bd8_0    conda-forge
     curl                      7.79.1               h2574ce0_1    conda-forge
     cusignal                  21.10.00        py37_gff14a10_0    rapidsai
     cuspatial                 21.10.00        py38_gba20298_0    rapidsai
     custreamz                 21.10.00        py38_g072fd862cc_0    rapidsai
     cuxfilter                 21.10.00        py38_g003d3d6_0    rapidsai
     cycler                    0.10.0                     py_2    conda-forge
     cyrus-sasl                2.1.27               h230043b_3    conda-forge
     cytoolz                   0.11.0           py38h497a2fe_3    conda-forge
     dash                      2.0.0                    pypi_0    pypi
     dash-auth                 1.4.1              pyhd3deb0d_0    conda-forge
     dash-bootstrap-components 0.13.1             pyhd8ed1ab_0    conda-forge
     dash-core-components      2.0.0                    pypi_0    pypi
     dash-cytoscape            0.3.0                    pypi_0    pypi
     dash-daq                  0.5.0              pyh9f0ad1d_1    conda-forge
     dash-extensions           0.0.60                   pypi_0    pypi
     dash-html-components      2.0.0                    pypi_0    pypi
     dash-renderer             1.9.1              pyhd8ed1ab_0    conda-forge
     dash-table                5.0.0                    pypi_0    pypi
     dash-uploader             0.6.0                    pypi_0    pypi
     dask                      2021.9.1           pyhd8ed1ab_0    conda-forge
     dask-core                 2021.9.1           pyhd8ed1ab_0    conda-forge
     dask-cuda                 21.10.00                 py38_0    rapidsai
     dask-cudf                 21.10.00        py38_g072fd862cc_0    rapidsai
     dataclasses               0.8                pyhc8e2a94_3    conda-forge
     datasets                  1.13.2             pyhd8ed1ab_0    conda-forge
     datashader                0.11.1             pyh9f0ad1d_0    conda-forge
     datashape                 0.5.4                      py_1    conda-forge
     debugpy                   1.4.1            py38h709712a_0    conda-forge
     decorator                 4.4.2                      py_0    conda-forge
     defusedxml                0.7.1              pyhd8ed1ab_0    conda-forge
     dill                      0.3.4              pyhd8ed1ab_0    conda-forge
     distributed               2021.9.1         py38h578d9bd_0    conda-forge
     dlpack                    0.5                  h9c3ff4c_0    conda-forge
     easynmt                   2.0.1                    pypi_0    pypi
     entrypoints               0.3             pyhd8ed1ab_1003    conda-forge
     expat                     2.4.1                h9c3ff4c_0    conda-forge
     faiss-proc                1.0.0                      cuda    rapidsai
     fastavro                  1.4.5            py38h497a2fe_0    conda-forge
     fastrlock                 0.6              py38h709712a_1    conda-forge
     fasttext                  0.9.2                    pypi_0    pypi
     filelock                  3.3.0              pyhd8ed1ab_0    conda-forge
     fiona                     1.8.20           py38hbb147eb_1    conda-forge
     flask                     2.0.2              pyhd8ed1ab_0    conda-forge
     flask-caching             1.10.1                   pypi_0    pypi
     flask-compress            1.10.1             pyhd8ed1ab_0    conda-forge
     flask-seasurf             0.3.1              pyhd8ed1ab_0    conda-forge
     font-ttf-dejavu-sans-mono 2.37                 hab24e00_0    conda-forge
     font-ttf-inconsolata      3.000                h77eed37_0    conda-forge
     font-ttf-source-code-pro  2.038                h77eed37_0    conda-forge
     font-ttf-ubuntu           0.83                 hab24e00_0    conda-forge
     fontconfig                2.13.1            hba837de_1005    conda-forge
     fonts-conda-ecosystem     1                             0    conda-forge
     fonts-conda-forge         1                             0    conda-forge
     freetype                  2.10.4               h0708190_1    conda-forge
     freexl                    1.0.6                h7f98852_0    conda-forge
     fsspec                    2021.10.0          pyhd8ed1ab_0    conda-forge
     future                    0.18.2           py38h578d9bd_3    conda-forge
     gdal                      3.3.1            py38h81a01a0_3    conda-forge
     geopandas                 0.9.0              pyhd8ed1ab_1    conda-forge
     geopandas-base            0.9.0              pyhd8ed1ab_1    conda-forge
     geos                      3.9.1                h9c3ff4c_2    conda-forge
     geotiff                   1.6.0                h4f31c25_6    conda-forge
     gettext                   0.19.8.1          h73d1719_1008    conda-forge
     gevent                    21.8.0           py38h497a2fe_0    conda-forge
     gflags                    2.2.2             he1b5a44_1004    conda-forge
     giflib                    5.2.1                h36c2ea0_2    conda-forge
     git                       2.33.0          pl5321hc30692c_1    conda-forge
     glog                      0.5.0                h48cff8f_0    conda-forge
     google-cloud-cpp          1.29.0               hb967e95_1    conda-forge
     gpuci-tools               0.3.1                        10    gpuci
     greenlet                  1.1.2            py38h709712a_0    conda-forge
     grpc-cpp                  1.39.1               h850795e_1    conda-forge
     hdbscan                   0.8.27           py38h5c078b8_0    conda-forge
     hdf4                      4.2.15               h10796ff_3    conda-forge
     hdf5                      1.12.1          nompi_h2750804_101    conda-forge
     heapdict                  1.0.1                      py_0    conda-forge
     huggingface_hub           0.0.19             pyhd8ed1ab_0    conda-forge
     humanize                  3.12.0                   pypi_0    pypi
     icu                       68.1                 h58526e2_0    conda-forge
     idna                      3.1                pyhd3deb0d_0    conda-forge
     imagecodecs               2021.7.30        py38hb5ce8f7_1    conda-forge
     imageio                   2.9.0                      py_0    conda-forge
     importlib-metadata        4.8.1            py38h578d9bd_0    conda-forge
     importlib_metadata        4.8.1                hd8ed1ab_0    conda-forge
     iniconfig                 1.1.1                    pypi_0    pypi
     ipykernel                 6.4.1            py38he5a9106_0    conda-forge
     ipython                   7.28.0           py38he5a9106_0    conda-forge
     ipython_genutils          0.2.0                      py_1    conda-forge
     ipywidgets                7.6.5              pyhd8ed1ab_0    conda-forge
     itsdangerous              2.0.1              pyhd8ed1ab_0    conda-forge
     jbig                      2.1               h7f98852_2003    conda-forge
     jedi                      0.18.0           py38h578d9bd_2    conda-forge
     jinja2                    3.0.1              pyhd8ed1ab_0    conda-forge
     joblib                    1.0.1              pyhd8ed1ab_0    conda-forge
     jpeg                      9d                   h36c2ea0_0    conda-forge
     jpype1                    1.3.0            py38h1fd1430_0    conda-forge
     json-c                    0.15                 h98cffda_0    conda-forge
     jsonschema                4.0.1              pyhd8ed1ab_0    conda-forge
     jupyter-server-proxy      3.1.0              pyhd8ed1ab_0    conda-forge
     jupyter_client            7.0.6              pyhd8ed1ab_0    conda-forge
     jupyter_core              4.8.1            py38h578d9bd_0    conda-forge
     jupyter_server            1.11.1             pyhd8ed1ab_0    conda-forge
     jupyterlab_pygments       0.1.2              pyh9f0ad1d_0    conda-forge
     jupyterlab_widgets        1.0.2              pyhd8ed1ab_0    conda-forge
     jxrlib                    1.1                  h7f98852_2    conda-forge
     kealib                    1.4.14               h87e4c3c_3    conda-forge
     kiwisolver                1.3.2            py38h1fd1430_0    conda-forge
     krb5                      1.19.2               hcc1bbae_2    conda-forge
     lcms2                     2.12                 hddcbb42_0    conda-forge
     ld_impl_linux-64          2.36.1               hea4e1c9_2    conda-forge
     lerc                      2.2.1                h9c3ff4c_0    conda-forge
     libaec                    1.0.6                h9c3ff4c_0    conda-forge
     libblas                   3.9.0                     8_mkl    conda-forge
     libbrotlicommon           1.0.9                h7f98852_5    conda-forge
     libbrotlidec              1.0.9                h7f98852_5    conda-forge
     libbrotlienc              1.0.9                h7f98852_5    conda-forge
     libcblas                  3.9.0                     8_mkl    conda-forge
     libcrc32c                 1.1.1                h9c3ff4c_2    conda-forge
     libcucim                  21.10.00        cuda11.0_gd7ac21f_0    rapidsai
     libcudf                   21.10.00        cuda11.0_g072fd862cc_0    rapidsai
     libcudf_kafka             21.10.00          g072fd862cc_0    rapidsai
     libcugraph                21.10.00        cuda11.0_g84617024_0    rapidsai
     libcuml                   21.10.00        cuda11.0_g0fd3503ba_0    rapidsai
     libcumlprims              21.10.00        cuda11.0_g167dc59_0    nvidia
     libcurl                   7.79.1               h2574ce0_1    conda-forge
     libcuspatial              21.10.00        cuda11.0_gba20298_0    rapidsai
     libdap4                   3.20.6               hd7c4107_2    conda-forge
     libdeflate                1.7                  h7f98852_5    conda-forge
     libedit                   3.1.20191231         he28a2e2_2    conda-forge
     libev                     4.33                 h516909a_1    conda-forge
     libevent                  2.1.10               h9b69904_4    conda-forge
     libfaiss                  1.7.0           cuda110h8045045_8_cuda    conda-forge
     libffi                    3.4.2                h9c3ff4c_4    conda-forge
     libgcc-ng                 9.4.0                hfa6338b_9    conda-forge
     libgcrypt                 1.9.4                h7f98852_0    conda-forge
     libgdal                   3.3.1                h6214c1d_3    conda-forge
     libgfortran-ng            11.2.0               h69a702a_9    conda-forge
     libgfortran5              11.2.0               h5c6108e_9    conda-forge
     libglib                   2.68.4               h174f98d_1    conda-forge
     libgpg-error              1.42                 h9c3ff4c_0    conda-forge
     libgsasl                  1.10.0               h5b4c23d_0    conda-forge
     libhwloc                  2.3.0                h5e5b7d1_1    conda-forge
     libiconv                  1.16                 h516909a_0    conda-forge
     libkml                    1.3.0             hd79254b_1012    conda-forge
     liblapack                 3.9.0                     8_mkl    conda-forge
     liblapacke                3.9.0                     8_mkl    conda-forge
     libllvm10                 10.0.1               he513fc3_3    conda-forge
     libnetcdf                 4.8.1           nompi_hb3fd0d9_101    conda-forge
     libnghttp2                1.43.0               h812cca2_1    conda-forge
     libntlm                   1.4               h7f98852_1002    conda-forge
     libopenblas               0.3.17          pthreads_h8fe5266_1    conda-forge
     libpng                    1.6.37               h21135ba_2    conda-forge
     libpq                     13.3                 hd57d9b9_0    conda-forge
     libprotobuf               3.16.0               h780b84a_0    conda-forge
     librdkafka                1.6.1                hc49e61c_1    conda-forge
     librmm                    21.10.00        cuda11.0_g72ee41f_0    rapidsai
     librttopo                 1.1.0                h1185371_6    conda-forge
     libsodium                 1.0.18               h36c2ea0_1    conda-forge
     libspatialindex           1.9.3                h9c3ff4c_4    conda-forge
     libspatialite             5.0.1                h8694cbe_6    conda-forge
     libssh2                   1.10.0               ha56f1ee_2    conda-forge
     libstdcxx-ng              9.4.0                h79bfe98_9    conda-forge
     libthrift                 0.14.2               he6d91bd_1    conda-forge
     libtiff                   4.3.0                hf544144_1    conda-forge
     libutf8proc               2.6.1                h7f98852_0    conda-forge
     libuuid                   2.32.1            h7f98852_1000    conda-forge
     libuv                     1.42.0               h7f98852_0    conda-forge
     libwebp                   1.2.1                h3452ae3_0    conda-forge
     libwebp-base              1.2.1                h7f98852_0    conda-forge
     libxcb                    1.13              h7f98852_1003    conda-forge
     libxgboost                1.4.2dev.rapidsai21.10      cuda11.0_0    rapidsai
     libxml2                   2.9.12               h72842e0_0    conda-forge
     libzip                    1.8.0                h4de3113_1    conda-forge
     libzlib                   1.2.11            h36c2ea0_1013    conda-forge
     libzopfli                 1.0.3                h9c3ff4c_0    conda-forge
     llvm-openmp               12.0.1               h4bd325d_1    conda-forge
     llvmlite                  0.36.0           py38h4630a5e_0    conda-forge
     locket                    0.2.0                      py_2    conda-forge
     lz4-c                     1.9.3                h9c3ff4c_1    conda-forge
     mapclassify               2.4.3              pyhd8ed1ab_0    conda-forge
     markdown                  3.3.4              pyhd8ed1ab_0    conda-forge
     markupsafe                2.0.1            py38h497a2fe_0    conda-forge
     matplotlib-base           3.4.3            py38hf4fb855_1    conda-forge
     matplotlib-inline         0.1.3              pyhd8ed1ab_0    conda-forge
     mistune                   0.8.4           py38h497a2fe_1004    conda-forge
     mkl                       2020.4             h726a3e6_304    conda-forge
     mkl-devel                 2020.4             ha770c72_305    conda-forge
     mkl-include               2020.4             h726a3e6_304    conda-forge
     more-itertools            8.10.0                   pypi_0    pypi
     msgpack-python            1.0.2            py38h1fd1430_1    conda-forge
     multidict                 5.2.0            py38h497a2fe_0    conda-forge
     multipledispatch          0.6.0                      py_0    conda-forge
     multiprocess              0.70.12.2        py38h497a2fe_0    conda-forge
     munch                     2.5.0                      py_0    conda-forge
     nbclient                  0.5.4              pyhd8ed1ab_0    conda-forge
     nbconvert                 6.2.0            py38h578d9bd_0    conda-forge
     nbformat                  5.1.3              pyhd8ed1ab_0    conda-forge
     nccl                      2.11.4.1             h96e36e3_0    conda-forge
     ncurses                   6.2                  h58526e2_4    conda-forge
     nest-asyncio              1.5.1              pyhd8ed1ab_0    conda-forge
     netifaces                 0.10.9          py38h497a2fe_1003    conda-forge
     networkx                  2.6.3              pyhd8ed1ab_0    conda-forge
     ninja                     1.10.2               h4bd325d_1    conda-forge
     nlohmann_json             3.9.1                h9c3ff4c_1    conda-forge
     nltk                      3.6.5              pyhd8ed1ab_0    conda-forge
     nodejs                    14.17.4              h92b4a50_0    conda-forge
     nostril                   1.2.0                    pypi_0    pypi
     notebook                  6.4.4              pyha770c72_0    conda-forge
     nspr                      4.30                 h9c3ff4c_0    conda-forge
     nss                       3.69                 hb5efdd6_1    conda-forge
     numba                     0.53.1           py38h8b71fd7_1    conda-forge
     numpy                     1.21.2           py38he2449b9_0    conda-forge
     nvtx                      0.2.3            py38h497a2fe_0    conda-forge
     olefile                   0.46               pyh9f0ad1d_1    conda-forge
     openjdk                   8.0.302              h7f98852_0    conda-forge
     openjpeg                  2.4.0                hb52868f_1    conda-forge
     openssl                   1.1.1l               h7f98852_0    conda-forge
     orc                       1.6.10               h58a87f1_0    conda-forge
     packaging                 21.0               pyhd8ed1ab_0    conda-forge
     pandas                    1.3.3            py38h43a58ef_0    conda-forge
     pandoc                    2.14.2               h7f98852_0    conda-forge
     pandocfilters             1.5.0              pyhd8ed1ab_0    conda-forge
     panel                     0.12.4             pyhd8ed1ab_0    conda-forge
     param                     1.11.1             pyh6c4a22f_0    conda-forge
     parquet-cpp               1.5.1                         2    conda-forge
     parso                     0.8.2              pyhd8ed1ab_0    conda-forge
     partd                     1.2.0              pyhd8ed1ab_0    conda-forge
     patsy                     0.5.2              pyhd8ed1ab_0    conda-forge
     pcre                      8.45                 h9c3ff4c_0    conda-forge
     pcre2                     10.37                h032f7d1_0    conda-forge
     perl                      5.32.1          1_h7f98852_perl5    conda-forge
     pexpect                   4.8.0              pyh9f0ad1d_2    conda-forge
     pickleshare               0.7.5                   py_1003    conda-forge
     pillow                    8.3.2            py38h8e6f84c_0    conda-forge
     pip                       21.3               pyhd8ed1ab_0    conda-forge
     pixman                    0.40.0               h36c2ea0_0    conda-forge
     plac                      1.3.3                    pypi_0    pypi
     plotly                    5.3.1              pyhd8ed1ab_0    conda-forge
     pluggy                    1.0.0                    pypi_0    pypi
     pooch                     1.5.1              pyhd8ed1ab_0    conda-forge
     poppler                   21.03.0              h93df280_0    conda-forge
     poppler-data              0.4.11               hd8ed1ab_0    conda-forge
     postgresql                13.3                 h2510834_0    conda-forge
     proj                      8.0.1                h277dcde_0    conda-forge
     prometheus_client         0.11.0             pyhd8ed1ab_0    conda-forge
     prompt-toolkit            3.0.20             pyha770c72_0    conda-forge
     protobuf                  3.16.0           py38h709712a_0    conda-forge
     psutil                    5.8.0            py38h497a2fe_1    conda-forge
     pthread-stubs             0.4               h36c2ea0_1001    conda-forge
     ptyprocess                0.7.0              pyhd3deb0d_0    conda-forge
     pure-sasl                 0.6.2              pyhd8ed1ab_0    conda-forge
     py                        1.10.0                   pypi_0    pypi
     py-xgboost                1.4.2dev.rapidsai21.10  cuda11.0py38_0    rapidsai
     pyarrow                   5.0.0           py38hed47224_4_cuda    conda-forge
     pybind11                  2.8.0                    pypi_0    pypi
     pycparser                 2.20               pyh9f0ad1d_2    conda-forge
     pyct                      0.4.6                      py_0    conda-forge
     pyct-core                 0.4.6                      py_0    conda-forge
     pydeck                    0.5.0              pyh9f0ad1d_0    conda-forge
     pyee                      8.1.0              pyh9f0ad1d_0    conda-forge
     pygments                  2.10.0             pyhd8ed1ab_0    conda-forge
     pyhive                    0.6.4              pyhd8ed1ab_0    conda-forge
     pynndescent               0.5.4              pyh6c4a22f_0    conda-forge
     pynvml                    11.0.0             pyhd8ed1ab_0    conda-forge
     pyopenssl                 21.0.0             pyhd8ed1ab_0    conda-forge
     pyparsing                 2.4.7              pyh9f0ad1d_0    conda-forge
     pyppeteer                 0.2.6              pyhd8ed1ab_0    conda-forge
     pyproj                    3.1.0            py38h4df08a6_4    conda-forge
     pyrsistent                0.17.3           py38h497a2fe_2    conda-forge
     pysocks                   1.7.1            py38h578d9bd_3    conda-forge
     pytest                    6.2.5                    pypi_0    pypi
     python                    3.8.12          hb7a2778_1_cpython    conda-forge
     python-confluent-kafka    1.6.0            py38h497a2fe_1    conda-forge
     python-dateutil           2.8.2              pyhd8ed1ab_0    conda-forge
     python-xxhash             2.0.2            py38h497a2fe_0    conda-forge
     python_abi                3.8                      2_cp38    conda-forge
     pytz                      2021.3             pyhd8ed1ab_0    conda-forge
     pyviz_comms               2.1.0              pyhd8ed1ab_0    conda-forge
     pywavelets                1.1.1            py38h6c62de6_3    conda-forge
     pyyaml                    5.4.1            py38h497a2fe_1    conda-forge
     pyzmq                     22.3.0           py38h2035c66_0    conda-forge
     rapids                    21.10.00        cuda11.0_py38_ge66f011_114    rapidsai
     rapids-blazing            21.10.00        cuda11.0_py38_ge66f011_114    rapidsai
     rapids-xgboost            21.10.00        cuda11.0_py38_ge66f011_114    rapidsai
     re2                       2021.09.01           h9c3ff4c_0    conda-forge
     readline                  8.1                  h46c0cb4_0    conda-forge
     redis-py                  3.5.3              pyh9f0ad1d_0    conda-forge
     regex                     2021.10.8        py38h497a2fe_0    conda-forge
     requests                  2.26.0             pyhd8ed1ab_0    conda-forge
     requests-unixsocket       0.2.0                      py_0    conda-forge
     retrying                  1.3.3                      py_2    conda-forge
     rmm                       21.10.00        cuda_11.0_py38_g72ee41f_0    rapidsai
     rtree                     0.9.7            py38h02d302b_2    conda-forge
     s2n                       1.0.10               h9b69904_0    conda-forge
     sacremoses                0.0.43             pyh9f0ad1d_0    conda-forge
     scikit-image              0.18.1           py38h51da96c_0    conda-forge
     scikit-learn              0.24.2           py38hacb3eff_1    conda-forge
     scipy                     1.7.1            py38h56a6a73_0    conda-forge
     send2trash                1.8.0              pyhd8ed1ab_0    conda-forge
     sentence-transformers     2.1.0                    pypi_0    pypi
     sentencepiece             0.1.96                   pypi_0    pypi
     setuptools                49.6.0           py38h578d9bd_3    conda-forge
     shapely                   1.7.1            py38hb7fe4a8_5    conda-forge
     simpervisor               0.4                pyhd8ed1ab_0    conda-forge
     six                       1.16.0             pyh6c4a22f_0    conda-forge
     snappy                    1.1.8                he1b5a44_3    conda-forge
     sniffio                   1.2.0            py38h578d9bd_1    conda-forge
     sortedcontainers          2.4.0              pyhd8ed1ab_0    conda-forge
     spdlog                    1.8.5                h4bd325d_0    conda-forge
     sqlalchemy                1.4.25           py38h497a2fe_0    conda-forge
     sqlite                    3.36.0               h9cd32fc_2    conda-forge
     statsmodels               0.13.0           py38h6c62de6_0    conda-forge
     streamz                   0.6.3              pyh6c4a22f_0    conda-forge
     tabulate                  0.8.9                    pypi_0    pypi
     tbb                       2020.2               h4bd325d_4    conda-forge
     tblib                     1.7.0              pyhd8ed1ab_0    conda-forge
     tenacity                  8.0.1              pyhd8ed1ab_0    conda-forge
     terminado                 0.12.1           py38h578d9bd_0    conda-forge
     testpath                  0.5.0              pyhd8ed1ab_0    conda-forge
     thai-segmenter            0.4.1                    pypi_0    pypi
     threadpoolctl             3.0.0              pyh8a188c0_0    conda-forge
     thrift                    0.14.0           py38h709712a_0    conda-forge
     thrift_sasl               0.4.3              pyhd8ed1ab_1    conda-forge
     tifffile                  2021.8.30          pyhd8ed1ab_0    conda-forge
     tiledb                    2.3.4                he87e0bf_0    conda-forge
     tk                        8.6.11               h27826a3_1    conda-forge
     tokenizers                0.10.3           py38hb63a372_1    conda-forge
     toml                      0.10.2                   pypi_0    pypi
     toolz                     0.11.1                     py_0    conda-forge
     torch                     1.9.1                    pypi_0    pypi
     torchvision               0.10.1                   pypi_0    pypi
     tornado                   6.1              py38h497a2fe_1    conda-forge
     tqdm                      4.62.3             pyhd8ed1ab_0    conda-forge
     traitlets                 5.1.0              pyhd8ed1ab_0    conda-forge
     transformers              4.11.3             pyhd8ed1ab_0    conda-forge
     treelite                  2.1.0            py38hdd725b4_0    conda-forge
     treelite-runtime          2.1.0                    pypi_0    pypi
     typing-extensions         3.10.0.2             hd8ed1ab_0    conda-forge
     typing_extensions         3.10.0.2           pyha770c72_0    conda-forge
     tzcode                    2021c                h7f98852_0    conda-forge
     tzdata                    2021c                he74cb21_0    conda-forge
     tzlocal                   2.0.0                      py_0    conda-forge
     ua-parser                 0.10.0             pyh9f0ad1d_0    conda-forge
     ucx                       1.11.1+gc58db6b      cuda11.0_0    rapidsai
     ucx-proc                  1.0.0                       gpu    rapidsai
     ucx-py                    0.22.0          py38_gc58db6b_0    rapidsai
     umap-learn                0.5.1            py38h578d9bd_1    conda-forge
     urllib3                   1.26.7             pyhd8ed1ab_0    conda-forge
     wcwidth                   0.2.5              pyh9f0ad1d_2    conda-forge
     webencodings              0.5.1                      py_1    conda-forge
     websocket-client          0.57.0           py38h578d9bd_4    conda-forge
     websockets                9.1              py38h497a2fe_0    conda-forge
     werkzeug                  2.0.1              pyhd8ed1ab_0    conda-forge
     wheel                     0.37.0             pyhd8ed1ab_1    conda-forge
     widgetsnbextension        3.5.1            py38h578d9bd_4    conda-forge
     xarray                    0.19.0             pyhd8ed1ab_1    conda-forge
     xerces-c                  3.2.3                h9d8b166_2    conda-forge
     xgboost                   1.4.2dev.rapidsai21.10  cuda11.0py38_0    rapidsai
     xorg-kbproto              1.0.7             h7f98852_1002    conda-forge
     xorg-libice               1.0.10               h7f98852_0    conda-forge
     xorg-libsm                1.2.3             hd9c2040_1000    conda-forge
     xorg-libx11               1.7.2                h7f98852_0    conda-forge
     xorg-libxau               1.0.9                h7f98852_0    conda-forge
     xorg-libxdmcp             1.1.3                h7f98852_0    conda-forge
     xorg-libxext              1.3.4                h7f98852_1    conda-forge
     xorg-libxrender           0.9.10            h7f98852_1003    conda-forge
     xorg-renderproto          0.11.1            h7f98852_1002    conda-forge
     xorg-xextproto            7.3.0             h7f98852_1002    conda-forge
     xorg-xproto               7.0.31            h7f98852_1007    conda-forge
     xxhash                    0.8.0                h7f98852_3    conda-forge
     xz                        5.2.5                h516909a_1    conda-forge
     yaml                      0.2.5                h516909a_0    conda-forge
     yarl                      1.6.3            py38h497a2fe_2    conda-forge
     zeromq                    4.3.4                h9c3ff4c_1    conda-forge
     zfp                       0.5.5                h9c3ff4c_7    conda-forge
     zict                      2.0.0                      py_0    conda-forge
     zipp                      3.6.0              pyhd8ed1ab_0    conda-forge
     zlib                      1.2.11            h36c2ea0_1013    conda-forge
     zope.event                4.5.0              pyh9f0ad1d_0    conda-forge
     zope.interface            5.4.0            py38h497a2fe_0    conda-forge
     zstd                      1.5.0                ha95c52a_0    conda-forge
alexwilson1 commented 3 years ago

Thank you for opening the issue. A minimal reproducer would be welcomed for a more informed answer. However, from available information, I see two possible causes. First, you may simply run out of memory (device or host memory). Or otherwise, the failure might happen because of an incompatibility between the CUDA version installed in the environment (11.6) and the one RAPIDS (and FAISS) was compiled with (11.2). You may be able to fix this by installing a 11.4 CUDA environment and RAPIDS software suite.

For the environment given above, the following reproducers should work:

from cuml.manifold.umap import UMAP
import numpy as np
umap_model = UMAP(n_components=4, n_neighbors=20, min_dist=0.1)
umap_model.fit_transform(np.random.rand(285658, 768))

Gives error: RuntimeError: Error in virtual void faiss::gpu::StandardGpuResourcesImpl::initializeForDevice(int) at /home/conda/feedstock_root/build_artifacts/faiss-split_1618468126454/work/faiss/gpu/StandardGpuResources.cpp:285: Error: 'err == cudaSuccess' failed: failed to cudaHostAlloc 268435456 bytes for CPU <-> GPU async copy buffer (error 2 out of memory)

Whereas:

from cuml.manifold.umap import UMAP
import numpy as np
umap_model = UMAP(n_components=4, n_neighbors=20, min_dist=0.1)
umap_model.fit_transform(np.random.rand(700000, 768))

Executes successfully, as does:

from cuml.manifold.umap import UMAP
import numpy as np
umap_model = UMAP(n_components=4, n_neighbors=20, min_dist=0.1)
umap_model.fit_transform(np.random.rand(100000, 768))

I can execute these in any order, and they pass/fail reliably (matrix of dim 285658, 768 always fails but 700000, 100000 always passes). I'm wondering if the FAISS code is taking a different path within this range that is causing OOM?

viclafargue commented 3 years ago

I'm wondering if the FAISS code is taking a different path within this range that is causing OOM?

It's probably the case. However, I could not reproduce the bug.

As @dantegd said, the driver should support all the CUDA features up to version 11.6. It shouldn't be the issue here. However, I think that the cudatoolkit package should match what the other RAPIDS packages were compiled with (11.2). Have you tried reproducing the bug on a clean Docker setup before installing PyTorch or any other package that might alter the cudatoolkit version?

alexwilson1 commented 3 years ago

I'm wondering if the FAISS code is taking a different path within this range that is causing OOM?

It's probably the case. However, I could not reproduce the bug.

As @dantegd said, the driver should support all the CUDA features up to version 11.6. It shouldn't be the issue here. However, I think that the cudatoolkit package should match what the other RAPIDS packages were compiled with (11.2). Have you tried reproducing the bug on a clean Docker setup before installing PyTorch or any other package that might alter the cudatoolkit version?

I can repro with the following steps:

Stop all containers

Start the Rapids container in interactive mode

docker run -i --gpus all rapidsai/rapidsai:21.10-cuda11.2-base-ubuntu18.04-py3.8

Open the CLI of the container using Docker Desktop and execute the following:

# ls
NVIDIA_Deep_Learning_Container_License.pdf  utils
# bash
(rapids) root@70b49e55437e:/rapids# python
Python 3.8.12 | packaged by conda-forge | (default, Sep 29 2021, 19:52:28)
[GCC 9.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from cuml.manifold.umap import UMAP
>>> import numpy as np
>>> umap_model = UMAP(n_components=4, n_neighbors=20, min_dist=0.1)
>>> umap_model.fit_transform(np.random.rand(285658, 768))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/conda/envs/rapids/lib/python3.8/site-packages/cuml/internals/api_decorators.py", line 549, in inner_set_get
    ret_val = func(*args, **kwargs)
  File "cuml/manifold/umap.pyx", line 659, in cuml.manifold.umap.UMAP.fit_transform
  File "/opt/conda/envs/rapids/lib/python3.8/site-packages/cuml/internals/api_decorators.py", line 409, in inner_with_setters
    return func(*args, **kwargs)
  File "cuml/manifold/umap.pyx", line 600, in cuml.manifold.umap.UMAP.fit
RuntimeError: Error in virtual void faiss::gpu::StandardGpuResourcesImpl::initializeForDevice(int) at /home/conda/feedstock_root/build_artifacts/faiss-split_1618468141526/work/faiss/gpu/StandardGpuResources.cpp:285: Error: 'err == cudaSuccess' failed: failed to cudaHostAlloc 268435456 bytes for CPU <-> GPU async copy buffer (error 2 out of memory)
>>>

Honestly I'm quite surprised, I expected it to be a dependency related issue.

Note, as before, all other containers are stopped and nothing is using the GPU (apart from Windows desktop manager), task manager shows 1.6GB/24GB memory used on the GPU so there should be more than enough(?). In addition, the system is only using 30GB of it's 128GB RAM.

viclafargue commented 3 years ago

I ran the same commands, but could not reproduce the error. It is a possibility that the problem may arise because of the Windows host (I have a Linux host). The cudaHostAlloc call allocates page-locked memory thanks to the mlock syscall in Linux. However, this call might not be appropriately transmitted to the host OS.

If you want to investigate this lead, you can try the followings:

alexwilson1 commented 3 years ago

Thanks for the reply!

  1. I'll try and deploy on a K8s Linux host later today.
  2. IPC_LOCK still resulted in the same error unfortunately
  3. I ran the ulimit -acommand and the result is: 
    (rapids) root@14605cd77a24:/rapids# ulimit -a
core file size          (blocks, -c) 0
data seg size           (kbytes, -d) unlimited
scheduling priority             (-e) 0
file size               (blocks, -f) unlimited
pending signals                 (-i) 257249
max locked memory       (kbytes, -l) 82000
max memory size         (kbytes, -m) unlimited
open files                      (-n) 1048576
pipe size            (512 bytes, -p) 8
POSIX message queues     (bytes, -q) 819200
real-time priority              (-r) 0
stack size              (kbytes, -s) 8192
cpu time               (seconds, -t) unlimited
max user processes              (-u) unlimited
virtual memory          (kbytes, -v) unlimited
file locks                      (-x) unlimited
  4. Trying to increase locked memory limit does not work (rapids) root@14605cd77a24:/rapids# ulimit -l 999999999999999 bash: ulimit: max locked memory: cannot modify limit: Operation not permitted

I'll spend time trying to deploy on a Linux host first before we try and figure out the locked memory issue.

alexwilson1 commented 3 years ago

Still waiting to try this on a Linux host, but could this issue be related to this

As you have said @viclafargue, it seems like allocating large blocks may be a possible limitation of WSL2...

viclafargue commented 3 years ago

Still waiting to try this on a Linux host, but could this issue be related to this

As you have said @viclafargue, it seems like allocating large blocks may be a possible limitation of WSL2...

Thank you for spotting this. It may very well be the issue indeed.

alexwilson1 commented 3 years ago

Still waiting to try this on a Linux host, but could this issue be related to this As you have said @viclafargue, it seems like allocating large blocks may be a possible limitation of WSL2...

Thank you for spotting this. It may very well be the issue indeed.

No problem! Got it set up on a Linux host and can't repro the issue, so it seems like this is the WSL2 limitation.

My WSL2 setup is:

# uname -r
5.10.60.1-microsoft-standard-WSL2

@viclafargue what would be the recommended course of action here please?

viclafargue commented 3 years ago

I could communicate internally on the matter. As the thread you pointed out suggested, there are indeed some limits on how much host memory can be pinned on WSL at once. CUDA on WSL is still in development.

alexwilson1 commented 3 years ago

Thank you @viclafargue! I'm glad we got to the bottom of this, please keep us updated.

github-actions[bot] commented 2 years ago

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

github-actions[bot] commented 2 years ago

This issue has been labeled inactive-90d due to no recent activity in the past 90 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.