You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
| PyTorch | Not applicable |`-DENABLE_PYTORCH=ON -DUSE_PT_PYTHON_LIBS=ON` plus the PyTorch CMake prefix |
44
+
| TensorFlow | Not applicable |`-DENABLE_TENSORFLOW=ON -DUSE_TF_PYTHON_LIBS=ON` and the selected Python executable |
45
+
| JAX with Python TensorFlow C++ libraries | Both TensorFlow roots are null |`-DENABLE_TENSORFLOW=ON -DUSE_TF_PYTHON_LIBS=ON` and the selected Python executable |
46
+
| JAX with external TensorFlow C++ | Only `tensorflow_root` is set |`-DENABLE_TENSORFLOW=ON -DUSE_TF_PYTHON_LIBS=OFF -DTENSORFLOW_ROOT=<tensorflow-root>`|
47
+
| JAX with TensorFlow C | Only `tensorflow_c_root` is set |`-DENABLE_JAX=ON -DCMAKE_PREFIX_PATH=<tensorflow-c-root>`|
48
+
| Paddle | Not applicable |`-DENABLE_PADDLE=ON -DPADDLE_INFERENCE_DIR=<paddle-inference-dir>`|
49
+
50
+
The Python TensorFlow C++ route requires the planned interpreter to import a
51
+
TensorFlow package that provides `libtensorflow_cc`. Record and install that
52
+
package during the Python gate before configuring CMake.
47
53
48
54
For PyTorch, discover the prefix from the planned interpreter in the same shell
0 commit comments