[Python] Add backward compatible for paddle2.4.2 (#1929)

* [Python] Add backward compatible for paddle2.4.2

* [Python] Add backward compatible for paddle2.4.2

* [scripts] update linux build scripts

python/fastdeploy/__init__.py

@@ -15,6 +15,7 @@ from __future__ import absolute_import
 import logging
 import os
 import sys
+import platform
 
 # Create a symbol link to tensorrt library.
 trt_directory = os.path.join(
@@ -42,23 +43,82 @@ if os.name != "nt" and os.path.exists(trt_directory):
                 break
     logging.basicConfig(level=logging.NOTSET)
 
+
+from .code_version import version, git_version, extra_version_info
+from .code_version import enable_trt_backend, enable_paddle_backend, with_gpu
+
 # Note(zhoushunjie): Fix the import order of paddle and fastdeploy library.
 # This solution will be removed it when the confilct of paddle and
 # fastdeploy is fixed.
+
+# Note(qiuyanjun): Add backward compatible for paddle 2.4.x
+sys_platform = platform.platform().lower()
+
+
+def get_paddle_version():
+  paddle_version = ""
+  try:
+    import pkg_resources
+    paddle_version = pkg_resources.require(
+      "paddlepaddle-gpu")[0].version.split(".post")[0]
+  except:
+    try:
+      paddle_version = pkg_resources.require(
+        "paddlepaddle")[0].version.split(".post")[0]
+    except:
+      pass
+  return paddle_version      
+
+
+def should_import_paddle():
+  if "paddle2.4" in extra_version_info:
+    paddle_version = get_paddle_version()
+    if paddle_version != "" and paddle_version <= '2.4.2' and paddle_version != "0.0.0":
+      return True
+  return False    
+
+
+def should_set_tensorrt():
+  if with_gpu == 'ON' and enable_paddle_backend == 'ON' and enable_trt_backend == 'ON':
+    return True
+  return False  
+
+
+def tensorrt_is_avaliable():
+  # Note(qiuyanjun): Only support linux now.
+  found_trt_lib = False
+  if ('linux' in sys_platform) and ('LD_LIBRARY_PATH' in os.environ.keys()):
+    for lib_path in os.environ['LD_LIBRARY_PATH'].split(':'):
+      if os.path.exists(os.path.join(lib_path, 'libnvinfer.so')):
+        found_trt_lib = True
+        break
+  return found_trt_lib 
+
+
 try:
-  # import paddle
-  import platform
-  sys_platform = platform.platform().lower()
   # windows: no conflict between fastdeploy and paddle.
   # linux: must import paddle first to solve the conflict.
   # macos: still can not solve the conflict between fastdeploy and paddle, 
   #        due to the global flags redefined in paddle/paddle_inference so.
   #        we got the error (ERROR: flag 'xxx' was defined more than once).
   if "linux" in sys_platform:
-    import paddle
+    if should_import_paddle():
+      import paddle # need import paddle first for paddle2.4.x
+      # check whether tensorrt in LD_LIBRARY_PATH for fastdeploy
+      if should_set_tensorrt() and (not tensorrt_is_avaliable()):
+        if os.path.exists(trt_directory):
+          logging.info(
+           "\n[WARNING] Can not find TensorRT lib in LD_LIBRARY_PATH for FastDeploy! \
+            \n[WARNING] Please export [ YOUR CUSTOM TensorRT ] lib path to LD_LIBRARY_PATH first, or run the command: \
+            \n[WARNING] Linux: 'export LD_LIBRARY_PATH=$(python -c 'from fastdeploy import trt_directory; print(trt_directory)'):$LD_LIBRARY_PATH'")
+        else:  
+          logging.info(
+           "\n[WARNING] Can not find TensorRT lib in LD_LIBRARY_PATH for FastDeploy! \
+            \n[WARNING] Please export [YOUR CUSTOM TensorRT] lib path to LD_LIBRARY_PATH first.")   
 except:
   pass
 
+
 from .c_lib_wrap import (
     ModelFormat,
     Backend,
@@ -92,5 +152,5 @@ from . import text
 from . import encryption
 from .download import download, download_and_decompress, download_model, get_model_list
 from . import serving
-from .code_version import version, git_version
+
 __version__ = version