paddlex 3.0.0rc1__py3-none-any.whl → 3.0.1__py3-none-any.whl

paddlex/repo_apis/PaddleOCR_api/text_rec/model.py

@@ -135,11 +135,9 @@ class TextRecModel(BaseModel):
         # PDX related settings
         device_type = device.split(":")[0]
         uniform_output_enabled = kwargs.pop("uniform_output_enabled", True)
-        export_with_pir = kwargs.pop("export_with_pir", False)
         config.update({"Global.uniform_output_enabled": uniform_output_enabled})
-        config.update({"Global.pdx_model_name": self.name})
-        if export_with_pir:
-            config.update({"Global.export_with_pir": export_with_pir})
+        config.update({"Global.model_name": self.name})
+        config.update({"Global.export_with_pir": kwargs.pop("export_with_pir", False)})
 
         self._assert_empty_kwargs(kwargs)
 
@@ -272,11 +270,9 @@ class TextRecModel(BaseModel):
 
         # PDX related settings
         uniform_output_enabled = kwargs.pop("uniform_output_enabled", True)
-        export_with_pir = kwargs.pop("export_with_pir", False)
         config.update({"Global.uniform_output_enabled": uniform_output_enabled})
-        config.update({"Global.pdx_model_name": self.name})
-        if export_with_pir:
-            config.update({"Global.export_with_pir": export_with_pir})
+        config.update({"Global.model_name": self.name})
+        config.update({"Global.export_with_pir": kwargs.pop("export_with_pir", False)})
 
         self._assert_empty_kwargs(kwargs)
 
@@ -301,7 +297,7 @@ class TextRecModel(BaseModel):
             save_dir (str, optional): the directory path to save output. Defaults to None.
 
         Returns:
-            CompletedProcess: the result of infering subprocess execution.
+            CompletedProcess: the result of inferring subprocess execution.
         """
         config = self.config.copy()
         cli_args = []

paddlex/repo_apis/PaddleOCR_api/text_rec/register.py

@@ -196,3 +196,21 @@ register_model_info(
         "supported_apis": ["train", "evaluate", "predict", "export", "infer"],
     }
 )
+
+register_model_info(
+    {
+        "model_name": "PP-OCRv5_server_rec",
+        "suite": "TextRec",
+        "config_path": osp.join(PDX_CONFIG_DIR, "PP-OCRv5_server_rec.yaml"),
+        "supported_apis": ["train", "evaluate", "predict", "export"],
+    }
+)
+
+register_model_info(
+    {
+        "model_name": "PP-OCRv5_mobile_rec",
+        "suite": "TextRec",
+        "config_path": osp.join(PDX_CONFIG_DIR, "PP-OCRv5_mobile_rec.yaml"),
+        "supported_apis": ["train", "evaluate", "predict", "export"],
+    }
+)

paddlex/repo_apis/PaddleOCR_api/text_rec/runner.py

@@ -131,7 +131,7 @@ class TextRecRunner(BaseRunner):
             device (str): unused.
 
         Returns:
-            CompletedProcess: the result of infering subprocess execution.
+            CompletedProcess: the result of inferring subprocess execution.
         """
         cmd = [self.python, "tools/infer/predict_rec.py", *cli_args]
         return self.run_cmd(cmd, switch_wdir=True, echo=True, silent=False)

paddlex/repo_apis/PaddleSeg_api/seg/model.py

@@ -388,7 +388,7 @@ class SegModel(BaseModel):
             save_dir (str, optional): the directory path to save output. Defaults to None.
 
         Returns:
-            CompletedProcess: the result of infering subprocess execution.
+            CompletedProcess: the result of inferring subprocess execution.
         """
         config = self.config.copy()
         cli_args = []

paddlex/repo_apis/PaddleSeg_api/seg/runner.py

@@ -151,7 +151,7 @@ class SegRunner(BaseRunner):
             device (str): unused.
 
         Returns:
-            CompletedProcess: the result of infering subprocess execution.
+            CompletedProcess: the result of inferring subprocess execution.
         """
         # `device` unused
         cli_args = self._gather_opts_args(cli_args)

paddlex/repo_apis/PaddleTS_api/ts_ad/config.py

@@ -24,7 +24,7 @@ class TSAnomalyConfig(BaseTSConfig):
 
     def update_input_len(self, seq_len: int):
         """
-        upadte the input sequence length
+        update the input sequence length
 
         Args:
             seq_len (int): input length
@@ -40,7 +40,7 @@ class TSAnomalyConfig(BaseTSConfig):
 
     def update_dataset(self, dataset_dir: str, dataset_type: str = None):
         """
-        upadte the dataset
+        update the dataset
 
         Args:
             dataset_dir (str): dataset root path
@@ -57,7 +57,7 @@ class TSAnomalyConfig(BaseTSConfig):
         update basic info including time_col, freq, target_cols.
 
         Args:
-            info_params (dict): upadte basic info
+            info_params (dict): update basic info
 
         Raises:
             TypeError: if info_params is not dict, raising TypeError

paddlex/repo_apis/PaddleTS_api/ts_cls/config.py

@@ -24,7 +24,7 @@ class TSClassifyConfig(BaseTSConfig):
 
     def update_dataset(self, dataset_dir: str, dataset_type: str = None):
         """
-        upadte the dataset
+        update the dataset
 
         Args:
             dataset_dir (str): dataset root path
@@ -41,7 +41,7 @@ class TSClassifyConfig(BaseTSConfig):
         update basic info including time_col, freq, target_cols.
 
         Args:
-            info_params (dict): upadte basic info
+            info_params (dict): update basic info
 
         Raises:
             TypeError: if info_params is not dict, raising TypeError

paddlex/repo_apis/PaddleTS_api/ts_fc/config.py

@@ -24,7 +24,7 @@ class LongForecastConfig(BaseTSConfig):
 
     def update_input_len(self, seq_len: int):
         """
-        upadte the input sequence length
+        update the input sequence length
 
         Args:
             seq_len (int): input length
@@ -72,7 +72,7 @@ class LongForecastConfig(BaseTSConfig):
 
     def update_dataset(self, dataset_dir: str, dataset_type: str = None):
         """
-        upadte the dataset
+        update the dataset
 
         Args:
             dataset_dir (str): dataset root path
@@ -89,7 +89,7 @@ class LongForecastConfig(BaseTSConfig):
         update basic info including time_col, freq, target_cols.
 
         Args:
-            info_params (dict): upadte basic info
+            info_params (dict): update basic info
 
         Raises:
             TypeError: if info_params is not dict, raising TypeError
@@ -104,7 +104,7 @@ class LongForecastConfig(BaseTSConfig):
         update patience.
 
         Args:
-            patience (int): upadte patience
+            patience (int): update patience
 
         Raises:
             RuntimeError: if patience is not found, raising RuntimeError

paddlex/repo_apis/PaddleVideo_api/video_cls/config.py

@@ -435,7 +435,7 @@ indicating that no pretrained model to be used."
         """update directory that save predicting output
 
         Args:
-            save_dir (str): the dicrectory path that save predicting output.
+            save_dir (str): the directory path that save predicting output.
         """
         self.update({"Infer.save_dir": save_dir})
 

paddlex/repo_apis/PaddleVideo_api/video_cls/model.py

@@ -257,7 +257,7 @@ class VideoClsModel(BaseModel):
             dict_path (str, optional): the label dict file path. Defaults to None.
 
         Returns:
-            CompletedProcess: the result of infering subprocess execution.
+            CompletedProcess: the result of inferring subprocess execution.
         """
         model_dir = abspath(model_dir)
         input_path = abspath(input_path)

paddlex/repo_apis/PaddleVideo_api/video_cls/runner.py

@@ -136,7 +136,7 @@ class VideoClsRunner(BaseRunner):
             device (str): unused.
 
         Returns:
-            CompletedProcess: the result of infering subprocess execution.
+            CompletedProcess: the result of inferring subprocess execution.
         """
         # `device` unused
         cmd = [self.python, "python/predict_cls.py", "-c", config_path, *cli_args]

paddlex/repo_apis/PaddleVideo_api/video_det/config.py

@@ -436,7 +436,7 @@ indicating that no pretrained model to be used."
         """update directory that save predicting output
 
         Args:
-            save_dir (str): the dicrectory path that save predicting output.
+            save_dir (str): the directory path that save predicting output.
         """
         self.update({"Infer.save_dir": save_dir})
 

paddlex/repo_apis/PaddleVideo_api/video_det/model.py

@@ -258,7 +258,7 @@ class VideoDetModel(BaseModel):
             dict_path (str, optional): the label dict file path. Defaults to None.
 
         Returns:
-            CompletedProcess: the result of infering subprocess execution.
+            CompletedProcess: the result of inferring subprocess execution.
         """
         model_dir = abspath(model_dir)
         input_path = abspath(input_path)

paddlex/repo_apis/PaddleVideo_api/video_det/runner.py

@@ -136,7 +136,7 @@ class VideoDetRunner(BaseRunner):
             device (str): unused.
 
         Returns:
-            CompletedProcess: the result of infering subprocess execution.
+            CompletedProcess: the result of inferring subprocess execution.
         """
         # `device` unused
         cmd = [self.python, "python/predict_cls.py", "-c", config_path, *cli_args]

paddlex/repo_apis/base/config.py

@@ -106,7 +106,7 @@ class BaseConfig(_Config, metaclass=abc.ABCMeta):
     """
     Abstract base class of Config.
 
-    Config provides the funtionality to load, parse, or dump to a configuration
+    Config provides the functionality to load, parse, or dump to a configuration
     file with a specific format. Also, it provides APIs to update configurations
     of several important hyperparameters and model components.
     """

paddlex/repo_manager/core.py

@@ -101,7 +101,7 @@ def setup(
     """setup"""
     if update_repos and use_local_repos:
         logging.error(
-            f"The `--update_repos` and `--use_local_repos` should not be True at the same time. They are global setting for all repos. `--update_repos` means that update all repos to sync with remote, and `--use_local_repos` means that don't update when local repo is exsting."
+            f"The `--update_repos` and `--use_local_repos` should not be True at the same time. They are global setting for all repos. `--update_repos` means that update all repos to sync with remote, and `--use_local_repos` means that don't update when local repo is existing."
         )
         raise Exception()
 
@@ -136,7 +136,7 @@ def setup(
                 except EOFError:
                     logging.warning(
                         "Unable to read from stdin. Please set `--use_local_repos` to \
-                        True or False to apply a global setting for using exsting or re-getting repos."
+                        True or False to apply a global setting for using existing or re-getting repos."
                     )
                     raise
                 remove_existing = remove_existing.lower() in ("y", "yes")
@@ -174,7 +174,7 @@ def setup(
                 uninstall_existing = uninstall_existing.lower() in ("y", "yes")
 
             if uninstall_existing:
-                repo.uninstall()
+                build_repo_group_installer(repo).uninstall()
                 repos_to_install.append(repo)
             else:
                 logging.warning(

paddlex/repo_manager/meta.py

@@ -69,12 +69,12 @@ REPO_META = {
     "PaddleOCR": {
         "git_path": "/PaddlePaddle/PaddleOCR.git",
         "platform": "github",
-        "branch": "release/2.10",
+        "branch": "main",
         "install_pkg": False,
         "pdx_pkg_name": "PaddleOCR_api",
         "extra_pkgs": [
             "ppstructure/kie/requirements.txt",
-            "docs/algorithm/formula_recognition/requirements.txt",
+            "docs/version2.x/algorithm/formula_recognition/requirements.txt",
         ],
         "path_env": "PADDLE_PDX_PADDLEOCR_PATH",
         "requires": ["PaddleNLP"],
@@ -165,6 +165,10 @@ REPO_META = {
     },
 }
 
+REPO_DIST_NAMES = {
+    item["dist_name"] for item in REPO_META.values() if "dist_name" in item
+}
+
 
 def get_repo_meta(repo_name):
     """get_repo_meta"""

paddlex/repo_manager/repo.py

@@ -28,7 +28,7 @@ from ..utils.install import (
     install_packages_from_requirements_file,
     uninstall_packages,
 )
-from .meta import REPO_DOWNLOAD_BASE, get_repo_meta
+from .meta import REPO_DIST_NAMES, REPO_DOWNLOAD_BASE, get_repo_meta
 from .utils import (
     fetch_repo_using_git,
     install_external_deps,
@@ -133,8 +133,10 @@ class PPRepository(object):
                 try:
                     pip_install_opts = ["--no-deps"]
                     if editable:
-                        pip_install_opts.append("-e")
-                    install_packages(["."], pip_install_opts=pip_install_opts)
+                        reqs = ["-e ."]
+                    else:
+                        reqs = ["."]
+                    install_packages(reqs, pip_install_opts=pip_install_opts)
                     install_external_deps(self.name, self.root_dir)
                 finally:
                     if clean:
@@ -142,14 +144,16 @@ class PPRepository(object):
                         tmp_build_dir = "build"
                         if osp.exists(tmp_build_dir):
                             shutil.rmtree(tmp_build_dir)
-        for e in self.meta.get("extra", []):
+        for e in self.meta.get("extra_pkgs", []):
             if isinstance(e, tuple):
                 with switch_working_dir(osp.join(self.root_dir, e[0])):
+                    pip_install_opts = ["--no-deps"]
+                    if e[3]:
+                        reqs = ["-e ."]
+                    else:
+                        reqs = ["."]
                     try:
-                        pip_install_opts = ["--no-deps"]
-                        if e[3]:
-                            pip_install_opts.append("-e")
-                        install_packages(["."], pip_install_opts=pip_install_opts)
+                        install_packages(reqs, pip_install_opts=pip_install_opts)
                     finally:
                         if clean:
                             tmp_build_dir = "build"
@@ -159,9 +163,9 @@ class PPRepository(object):
     def uninstall_packages(self):
         """uninstall_packages"""
         pkgs = []
-        if self.install_pkg:
+        if self.meta["install_pkg"]:
             pkgs.append(self.dist_name)
-        for e in self.meta.get("extra", []):
+        for e in self.meta.get("extra_pkgs", []):
             if isinstance(e, tuple):
                 pkgs.append(e[1])
         uninstall_packages(pkgs)
@@ -206,7 +210,7 @@ class PPRepository(object):
         """get_deps"""
         # Merge requirement files
         req_list = [self.main_req_file]
-        for e in self.meta.get("extra", []):
+        for e in self.meta.get("extra_pkgs", []):
             if isinstance(e, tuple):
                 e = e[2] or osp.join(e[0], "requirements.txt")
             req_list.append(e)
@@ -327,7 +331,7 @@ class RepositoryGroupInstaller(object):
 
     def _sort_repos(self, repos, check_missing=False):
         # We sort the repos to ensure that the dependencies precede the
-        # dependant in the list.
+        # dependent in the list.
         name_meta_pairs = []
         for repo in repos:
             name_meta_pairs.append((repo.name, repo.meta))
@@ -358,9 +362,6 @@ class RepositoryGroupInstaller(object):
         return sorted_repos
 
     def _normalize_deps(self, deps, headline=None):
-        repo_pkgs = set(
-            repo.dist_name for repo in self.repos if repo.dist_name is not None
-        )
         lines = []
         if headline is not None:
             lines.append(headline)
@@ -376,7 +377,7 @@ class RepositoryGroupInstaller(object):
             # If `line` is not an empty line or a comment, it must be a requirement specifier.
             # Other forms may cause a parse error.
             req = Requirement(line_s)
-            if req.name in repo_pkgs:
+            if req.name in REPO_DIST_NAMES:
                 # Skip repo packages
                 continue
             elif req.name.replace("_", "-") in (

paddlex/utils/custom_device_list.py

@@ -135,8 +135,6 @@ NPU_BLACKLIST = [
     "whisper_base",
     "whisper_tiny",
     "UniMERNet",
-    "PP-FormulaNet-L",
-    "PP-FormulaNet-S",
 ]
 
 XPU_WHITELIST = [
@@ -191,6 +189,32 @@ XPU_WHITELIST = [
     "RLinear",
     "NLinear",
     "PicoDet_LCNet_x2_5_face",
+    "PP-LCNet_x1_0_doc_ori",
+    "UVDoc",
+    "PP-DocLayout-L",
+    "PP-DocLayout-M",
+    "PP-DocLayout-S",
+    "PP-DocLayout_plus-L",
+    "PP-DocBlockLayout",
+    "SLANeXt_wired",
+    "SLANeXt_wireless",
+    "PP-LCNet_x1_0_table_cls",
+    "RT-DETR-L_wired_table_cell_det",
+    "RT-DETR-L_wireless_table_cell_det",
+    "PP-OCRv4_server_seal_det",
+    "PP-OCRv4_mobile_seal_det",
+    "PP-OCRv5_server_rec",
+    "PP-OCRv5_mobile_rec",
+    "PPDocBee-2B",
+    "PPDocBee-7B",
+    "PP-FormulaNet-S",
+    "PP-FormulaNet-L",
+    "PP-FormulaNet-M",
+    "PP-FormulaNet_plus-M",
+    "PP-FormulaNet_plus-L",
+    "PP-OCRv5_server_det",
+    "PP-OCRv5_mobile_det",
+    "PP-Chart2Table",
 ]
 
 GCU_WHITELIST = [

paddlex/utils/deps.py

@@ -246,4 +246,4 @@ def require_paddle2onnx_plugin():
 
 
 def get_paddle2onnx_spec():
-    return "paddle2onnx >= 2.0.1"
+    return "paddle2onnx == 2.0.2rc3"

paddlex/utils/device.py

@@ -41,18 +41,25 @@ def constr_device(device_type, device_ids):
 
 
 def get_default_device():
-    avail_gpus = GPUtil.getAvailable()
-    if not avail_gpus:
-        # maybe edge devices like Jetson
+    try:
+        gpu_list = GPUtil.getGPUs()
+    except Exception:
+        logging.debug(
+            "Failed to query GPU devices. Falling back to CPU.", exc_info=True
+        )
+        has_gpus = False
+    else:
+        has_gpus = bool(gpu_list)
+    if not has_gpus:
+        # HACK
         if os.path.exists("/etc/nv_tegra_release"):
-            avail_gpus = [0]
-            logging.info(
-                "Detected that the current device is a Jetson edge device. The default behavior will be to use GPU: 0"
+            logging.debug(
+                "The current device appears to be an NVIDIA Jetson. GPU 0 will be used as the default device."
             )
-    if not avail_gpus:
+    if not has_gpus:
         return "cpu"
     else:
-        return constr_device("gpu", [avail_gpus[0]])
+        return constr_device("gpu", [0])
 
 
 def parse_device(device):

paddlex/utils/env.py

@@ -41,6 +41,8 @@ def get_paddle_cuda_version():
     import paddle.version
 
     cuda_version = paddle.version.cuda()
+    if cuda_version == "False":
+        return None
     return tuple(map(int, cuda_version.split(".")))
 
 
@@ -48,6 +50,8 @@ def get_paddle_cudnn_version():
     import paddle.version
 
     cudnn_version = paddle.version.cudnn()
+    if cudnn_version == "False":
+        return None
     return tuple(map(int, cudnn_version.split(".")))
 
 

paddlex/utils/flags.py

@@ -24,7 +24,6 @@ __all__ = [
     "INFER_BENCHMARK_ITERS",
     "INFER_BENCHMARK_WARMUP",
     "INFER_BENCHMARK_OUTPUT_DIR",
-    "INFER_BENCHMARK_USE_NEW_INFER_API",
     "FLAGS_json_format_model",
     "USE_PIR_TRT",
     "DISABLE_DEV_MODEL_WL",
@@ -52,6 +51,8 @@ FLAGS_json_format_model = get_flag_from_env_var("FLAGS_json_format_model", True)
 USE_PIR_TRT = get_flag_from_env_var("PADDLE_PDX_USE_PIR_TRT", True)
 DISABLE_DEV_MODEL_WL = get_flag_from_env_var("PADDLE_PDX_DISABLE_DEV_MODEL_WL", False)
 DISABLE_CINN_MODEL_WL = get_flag_from_env_var("PADDLE_PDX_DISABLE_CINN_MODEL_WL", False)
+LOCAL_FONT_FILE_PATH = get_flag_from_env_var("PADDLE_PDX_LOCAL_FONT_FILE_PATH", None)
+
 
 # Inference Benchmark
 INFER_BENCHMARK = get_flag_from_env_var("PADDLE_PDX_INFER_BENCHMARK", False)
@@ -67,6 +68,3 @@ INFER_BENCHMARK_ITERS = get_flag_from_env_var(
 INFER_BENCHMARK_USE_CACHE_FOR_READ = get_flag_from_env_var(
     "PADDLE_PDX_INFER_BENCHMARK_USE_CACHE_FOR_READ", False
 )
-INFER_BENCHMARK_USE_NEW_INFER_API = get_flag_from_env_var(
-    "PADDLE_PDX_INFER_BENCHMARK_USE_NEW_INFER_API", False
-)

paddlex/utils/fonts/__init__.py

@@ -18,7 +18,9 @@ import PIL
 from PIL import ImageFont
 
 from .. import logging
+from ..cache import CACHE_DIR
 from ..download import download
+from ..flags import LOCAL_FONT_FILE_PATH
 
 
 def get_font_file_path(file_name: str) -> str:
@@ -28,8 +30,8 @@ def get_font_file_path(file_name: str) -> str:
     Returns:
     str: The path to the font file.
     """
-    font_path = (Path(__file__).parent / file_name).resolve().as_posix()
-    if not Path(font_path).exists():
+    font_path = (Path(CACHE_DIR) / "fonts" / file_name).resolve().as_posix()
+    if not Path(font_path).is_file():
         download(
             url=f"https://paddle-model-ecology.bj.bcebos.com/paddlex/PaddleX3.0/fonts/{file_name}",
             save_path=font_path,
@@ -64,5 +66,33 @@ def create_font(txt: str, sz: tuple, font_path: str) -> ImageFont:
     return font
 
 
-PINGFANG_FONT_FILE_PATH = get_font_file_path("PingFang-SC-Regular.ttf")
-SIMFANG_FONT_FILE_PATH = get_font_file_path("simfang.ttf")
+def create_font_vertical(
+    txt: str, sz: tuple, font_path: str, scale=1.2
+) -> ImageFont.FreeTypeFont:
+    n = len(txt) if len(txt) > 0 else 1
+    base_font_size = int(sz[1] / n * 0.8 * scale)
+    base_font_size = max(base_font_size, 10)
+    font = ImageFont.truetype(font_path, base_font_size, encoding="utf-8")
+
+    if int(PIL.__version__.split(".")[0]) < 10:
+        max_char_width = max([font.getsize(c)[0] for c in txt])
+    else:
+        max_char_width = max([font.getlength(c) for c in txt])
+
+    if max_char_width > sz[0]:
+        new_size = int(base_font_size * sz[0] / max_char_width)
+        new_size = max(new_size, 10)
+        font = ImageFont.truetype(font_path, new_size, encoding="utf-8")
+
+    return font
+
+
+if Path(str(LOCAL_FONT_FILE_PATH)).is_file():
+    logging.warning(
+        f"Using the local font file(`{LOCAL_FONT_FILE_PATH}`) specified by `LOCAL_FONT_FILE_PATH`!"
+    )
+    PINGFANG_FONT_FILE_PATH = LOCAL_FONT_FILE_PATH
+    SIMFANG_FONT_FILE_PATH = LOCAL_FONT_FILE_PATH
+else:
+    PINGFANG_FONT_FILE_PATH = get_font_file_path("PingFang-SC-Regular.ttf")
+    SIMFANG_FONT_FILE_PATH = get_font_file_path("simfang.ttf")

paddlex/utils/misc.py

@@ -70,7 +70,7 @@ class CachedProperty(object):
     https://github.com/pydanny/cached-property/blob/master/cached_property.py .
 
     Note that this implementation does NOT work in multi-thread or coroutine
-    senarios.
+    scenarios.
     """
 
     def __init__(self, func):