Tensorflow is fully updated to TF2. Its official website shows that the old version of tensorflow 1.15 only supports Python 3.7 and cuda10, and the official apt source of the new version of Ubuntu does not have Python 3.7 at all.
If you want to use tf1.15 in the newer version of Python 3.8/cuda 11, one way is to recompile or download the binary WHL compiled by others. But recompiling and finding WHL are more troublesome. In order to make new graphics card users use tf1.15, Lao Huang cooperated with Google to maintain a tf1.15 library. Another method is to use docker, but docker is a little troublesome.
The address of the library is:https://github.com/NVIDIA/ten…
#It is recommended to operate in a virtual environment (not necessary) # python3 -m virtualenv venv # source venv/bin/activate pip install --upgrade pip pip install nvidia-pyindex pip install nvidia-tensorflow[horovod] pip install nvidia-tensorboard==1.15
import tensorflow as tf import tensorboard tf.enable_eager_execution() a = tf.random.uniform([1000, 1000]) b = tf.random.uniform([1000, 1000]) tf.matmul(a, b)
Check that the output is normal