I’m trying to build tensorflow locally, because that was given as a suggestion for resolving the error resulting from installing via pip.
I’ll list the error from the pip install first, then the compile error.
After doing:
python -m venv tf
source ./tf/bin/activate
pip install ‘tensorflow[and-cuda]’
the following:
python3 -c “import tensorflow as tf; print(tf.config.list_physical_devices(‘GPU’))”
gave me the error:
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
I0000 00:00:1776052480.694788 9230 port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variableTF_ENABLE_ONEDNN_OPTS=0.
I0000 00:00:1776052480.727001 9230 cpu_feature_guard.cc:227] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
I0000 00:00:1776052481.662388 9230 port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variableTF_ENABLE_ONEDNN_OPTS=0.
W0000 00:00:1776052481.929592 9230 gpu_device.cc:2365] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at https://www.tensorflow.org/install/gpu for how to download and setup the required libraries for your platform.
Skipping registering GPU devices…
I don’t understand the different messages about “9230 cpu_feature_guard.cc:227]” and “9230 gpu_device.cc:2365] Cannot dlopen some GPU libraries” … but the result is that it doesn’t find the GPU.
for some alternate validation of the python environment, torch successfully finds “1” GPU.
nvidia-smi reports:
NVIDIA-SMI 595.58.03 Driver Version: 595.58.03 CUDA Version: 13.2
Hoping that compiling locally would correctly discover CUDA, cudNN and thus find the GPU when invoked, I installed bazil and the build gives the following error:
ERROR: Error computing the main repository mapping: at /Storage/Projects/ML-LLMs-and-CUDA/tensorflow/tensorflow/workspace2.bzl:52:6: cannot load ‘@@xla//tools/def_file_filter:def_file_filter_configure.bzl’: no such file
any help/suggestions would be great!