feat(jax): add options to use TensorFlow C library to build the JAX backend

doc/install/easy-install.md

@@ -206,6 +206,10 @@ pip install deepmd-kit[jax]
 
 :::::
 
+To generate a SavedModel and use [the LAMMPS module](../third-party/lammps-command.md) and [the i-PI driver](../third-party/ipi.md),
+you need to install the TensorFlow.
+Switch to the TensorFlow {{ tensorflow_icon }} tab for more information.
+
 ::::::
 
 :::::::

doc/install/install-from-c-library.md

@@ -1,7 +1,7 @@
-# Install from pre-compiled C library {{ tensorflow_icon }}
+# Install from pre-compiled C library {{ tensorflow_icon }}, JAX {{ jax_icon }}
 
 :::{note}
-**Supported backends**: TensorFlow {{ tensorflow_icon }}
+**Supported backends**: TensorFlow {{ tensorflow_icon }}, JAX {{ jax_icon }}
 :::
 
 DeePMD-kit provides pre-compiled C library package (`libdeepmd_c.tar.gz`) in each [release](https://github.com/deepmodeling/deepmd-kit/releases). It can be used to build the [LAMMPS plugin](./install-lammps.md) and [GROMACS patch](./install-gromacs.md), as well as many [third-party software packages](../third-party/out-of-deepmd-kit.md), without building TensorFlow and DeePMD-kit on one's own.

doc/install/install-from-source.md

@@ -316,6 +316,15 @@ You can also download libtorch prebuilt library from the [PyTorch website](https
 
 :::
 
+:::{tab-item} JAX {{ jax_icon }}
+
+The JAX backend only depends on the TensorFlow C API, which is included in both TensorFlow C++ library and [TensorFlow C library](https://www.tensorflow.org/install/lang_c).
+If you want to use the TensorFlow C++ library, just enable the TensorFlow backend (which depends on the TensorFlow C++ library) and nothing else needs to do.
+If you want to use the TensorFlow C library and disable the TensorFlow backend,
+download the TensorFlow C library from [this page](https://www.tensorflow.org/install/lang_c#download_and_extract).
+
+:::
+
 ::::
 
 ### Install DeePMD-kit's C++ interface
@@ -369,6 +378,17 @@ cmake -DENABLE_PYTORCH=TRUE -DUSE_PT_PYTHON_LIBS=TRUE -DCMAKE_INSTALL_PREFIX=$de
 
 :::
 
+:::{tab-item} JAX {{ jax_icon }}
+
+If you want to use the TensorFlow C++ library, just enable the TensorFlow backend and nothing else needs to do.
+If you want to use the TensorFlow C library and disable the TensorFlow backend, set {cmake:variable}`ENABLE_JAX` to `ON` and `CMAKE_PREFIX_PATH` to the root directory of the [TensorFlow C library](https://www.tensorflow.org/install/lang_c).
+
+```bash
+cmake -DENABLE_JAX=ON -D CMAKE_PREFIX_PATH=${tensorflow_c_root} ..
+```
+
+:::
+
 ::::
 
 One may add the following CMake variables to `cmake` using the [`-D <var>=<value>` option](https://cmake.org/cmake/help/latest/manual/cmake.1.html#cmdoption-cmake-D):
@@ -378,6 +398,7 @@ One may add the following CMake variables to `cmake` using the [`-D <var>=<value
 **Type**: `BOOL` (`ON`/`OFF`), Default: `OFF`
 
 {{ tensorflow_icon }} {{ jax_icon }} Whether building the TensorFlow backend and the JAX backend.
+Setting this option to `ON` will also set {cmake:variable}`ENABLE_JAX` to `ON`.
 
 :::
 
@@ -389,6 +410,16 @@ One may add the following CMake variables to `cmake` using the [`-D <var>=<value
 
 :::
 
+:::{cmake:variable} ENABLE_JAX
+
+**Type**: `BOOL` (`ON`/`OFF`), Default: `OFF`
+
+{{ jax_icon }} Build the JAX backend.
+If {cmake:variable}`ENABLE_TENSORFLOW` is `ON`, the TensorFlow C++ library is used to build the JAX backend;
+If {cmake:variable}`ENABLE_TENSORFLOW` is `OFF`, the TensorFlow C library is used to build the JAX backend.
+
+:::
+
 :::{cmake:variable} TENSORFLOW_ROOT
 
 **Type**: `PATH`

source/CMakeLists.txt

@@ -4,6 +4,11 @@ project(DeePMD)
 
 option(ENABLE_TENSORFLOW "Enable TensorFlow interface" OFF)
 option(ENABLE_PYTORCH "Enable PyTorch interface" OFF)
+option(ENABLE_JAX "Enable JAX interface" OFF)
+if(ENABLE_TENSORFLOW)
+  # JAX requires TF C interface, contained in TF C++ library
+  set(ENABLE_JAX ON)
+endif()
 option(BUILD_TESTING "Build test and enable coverage" OFF)
 set(DEEPMD_C_ROOT
     ""
@@ -246,6 +251,22 @@ if(ENABLE_PYTORCH AND NOT DEEPMD_C_ROOT)
   list(APPEND BACKEND_LIBRARY_PATH ${PyTorch_LIBRARY_PATH})
   list(APPEND BACKEND_INCLUDE_DIRS ${TORCH_INCLUDE_DIRS})
 endif()
+if(ENABLE_JAX
+   AND BUILD_CPP_IF
+   AND NOT DEEPMD_C_ROOT)
+  # no way to find it using Python
+  find_package(TensorFlowC REQUIRED MODULE)
+  if(DEFINED TENSORFLOWC_LIBRARY)
+    list(APPEND BACKEND_LIBRARY_PATH ${TENSORFLOWC_LIBRARY})
+  endif()
+  if(DEFINED TENSORFLOWC_INCLUDE_DIR)
+    list(APPEND BACKEND_INCLUDE_DIRS ${TENSORFLOWC_INCLUDE_DIR})
+  endif()
+endif()
+if(NOT DEFINED OP_CXX_ABI)
+  # prevent setting an empty value; this is default on GCC>=5
+  set(OP_CXX_ABI 1)
+endif()
 # log enabled backends
 if(NOT DEEPMD_C_ROOT)
   message(STATUS "Enabled backends:")
@@ -255,8 +276,12 @@ if(NOT DEEPMD_C_ROOT)
   if(ENABLE_PYTORCH)
     message(STATUS "- PyTorch")
   endif()
+  if(ENABLE_JAX)
+    message(STATUS "- JAX")
+  endif()
   if(NOT ENABLE_TENSORFLOW
      AND NOT ENABLE_PYTORCH
+     AND NOT ENABLE_JAX
      AND NOT BUILD_PY_IF)
     message(FATAL_ERROR "No backend is enabled.")
   endif()

source/api_cc/CMakeLists.txt

@@ -23,6 +23,10 @@ if(ENABLE_PYTORCH
   target_link_libraries(${libname} PRIVATE "${TORCH_LIBRARIES}")
   target_compile_definitions(${libname} PRIVATE BUILD_PYTORCH)
 endif()
+if(ENABLE_JAX)
+  target_link_libraries(${libname} PRIVATE TensorFlow::tensorflow_c)
+  target_compile_definitions(${libname} PRIVATE BUILD_JAX)
+endif()
 
 target_include_directories(
   ${libname}

source/api_cc/src/DeepPot.cc

@@ -7,12 +7,14 @@
 #include "AtomMap.h"
 #include "common.h"
 #ifdef BUILD_TENSORFLOW
-#include "DeepPotJAX.h"
 #include "DeepPotTF.h"
 #endif
 #ifdef BUILD_PYTORCH
 #include "DeepPotPT.h"
 #endif
+#if defined(BUILD_TENSORFLOW) || defined(BUILD_JAX)
+#include "DeepPotJAX.h"
+#endif
 #include "device.h"
 
 using namespace deepmd;
@@ -63,7 +65,7 @@ void DeepPot::init(const std::string& model,
   } else if (deepmd::DPBackend::Paddle == backend) {
     throw deepmd::deepmd_exception("PaddlePaddle backend is not supported yet");
   } else if (deepmd::DPBackend::JAX == backend) {
-#ifdef BUILD_TENSORFLOW
+#if defined(BUILD_TENSORFLOW) || defined(BUILD_JAX)
     dp = std::make_shared<deepmd::DeepPotJAX>(model, gpu_rank, file_content);
 #else
     throw deepmd::deepmd_exception(

source/api_cc/src/DeepPotJAX.cc

@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: LGPL-3.0-or-later
-#ifdef BUILD_TENSORFLOW
+#if defined(BUILD_TENSORFLOW) || defined(BUILD_JAX)
 
 #include "DeepPotJAX.h"
 

source/cmake/FindTensorFlowC.cmake

@@ -0,0 +1,40 @@
+# Find TensorFlow C library (libtensorflow) Define target
+# TensorFlow::tensorflow_c If TensorFlow::tensorflow_cc is not found, also
+# define: - TENSORFLOWC_INCLUDE_DIR - TENSORFLOWC_LIBRARY
+
+if(TARGET TensorFlow::tensorflow_cc)
+  # since tensorflow_cc contain tensorflow_c, just use it
+  add_library(TensorFlow::tensorflow_c ALIAS TensorFlow::tensorflow_cc)
+  set(TensorFlowC_FOUND TRUE)
+endif()
+
+if(NOT TensorFlowC_FOUND)
+  find_path(
+    TENSORFLOWC_INCLUDE_DIR
+    NAMES tensorflow/c/c_api.h
+    PATH_SUFFIXES include
+    DOC "Path to TensorFlow C include directory")
+
+  find_library(
+    TENSORFLOWC_LIBRARY
+    NAMES tensorflow
+    PATH_SUFFIXES lib
+    DOC "Path to TensorFlow C library")
+
+  include(FindPackageHandleStandardArgs)
+  find_package_handle_standard_args(
+    TensorFlowC REQUIRED_VARS TENSORFLOWC_LIBRARY TENSORFLOWC_INCLUDE_DIR)
+
+  if(TensorFlowC_FOUND)
+    set(TensorFlowC_INCLUDE_DIRS ${TENSORFLOWC_INCLUDE_DIR})
+    set(TensorFlowC_LIBRARIES ${TENSORFLOWC_LIBRARY})
+  endif()
+
+  add_library(TensorFlow::tensorflow_c SHARED IMPORTED GLOBAL)
+  set_property(TARGET TensorFlow::tensorflow_c PROPERTY IMPORTED_LOCATION
+                                                        ${TENSORFLOWC_LIBRARY})
+  target_include_directories(TensorFlow::tensorflow_c
+                             INTERFACE ${TENSORFLOWC_INCLUDE_DIR})
+
+  mark_as_advanced(TENSORFLOWC_LIBRARY TENSORFLOWC_INCLUDE_DIR)
+endif()