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

paddlex/inference/pipelines/table_recognition/result.py

@@ -120,7 +120,7 @@ class TableRecognitionResult(BaseCVResult, HtmlMixin, XlsxMixin):
 
         if len(self["table_res_list"]) > 0:
             table_cell_img = Image.fromarray(
-                copy.deepcopy(self["doc_preprocessor_res"]["output_img"])
+                copy.deepcopy(self["doc_preprocessor_res"]["output_img"][:, :, ::-1])
             )
             table_draw = ImageDraw.Draw(table_cell_img)
             rectangle_color = (255, 0, 0)

paddlex/inference/pipelines/table_recognition/table_recognition_post_processing_v2.py

@@ -131,8 +131,8 @@ def compute_inter(rec1, rec2):
     Returns:
         float: Intersection over rec2_area
     """
-    x1_1, y1_1, x2_1, y2_1 = rec1
-    x1_2, y1_2, x2_2, y2_2 = rec2
+    x1_1, y1_1, x2_1, y2_1 = map(float, rec1)
+    x1_2, y1_2, x2_2, y2_2 = map(float, rec2)
     x_left = max(x1_1, x1_2)
     y_top = max(y1_1, y1_2)
     x_right = min(x2_1, x2_2)
@@ -413,8 +413,10 @@ def get_table_recognition_res(
     table_structure_result: list,
     table_cells_result: list,
     overall_ocr_res: OCRResult,
+    table_ocr_pred: dict,
     cells_texts_list: list,
     use_table_cells_ocr_results: bool,
+    use_table_cells_split_ocr: bool,
 ) -> SingleTableRecognitionResult:
     """
     Retrieve table recognition result from cropped image info, table structure prediction, and overall OCR result.
@@ -424,6 +426,7 @@ def get_table_recognition_res(
         table_structure_result (list): Predicted table structure.
         table_cells_result (list): Predicted table cells.
         overall_ocr_res (OCRResult): Overall OCR result from the input image.
+        table_ocr_pred (dict): Table OCR result from the input image.
         cells_texts_list (list): OCR results with cells.
         use_table_cells_ocr_results (bool): whether to use OCR results with cells.
 
@@ -432,9 +435,10 @@ def get_table_recognition_res(
     """
 
     table_cells_result = convert_to_four_point_coordinates(table_cells_result)
-
     table_box = np.array([table_box])
-    table_ocr_pred = get_sub_regions_ocr_res(overall_ocr_res, table_box)
+
+    if not (use_table_cells_ocr_results == True and use_table_cells_split_ocr == True):
+        table_ocr_pred = get_sub_regions_ocr_res(overall_ocr_res, table_box)
 
     crop_start_point = [table_box[0][0], table_box[0][1]]
     img_shape = overall_ocr_res["doc_preprocessor_res"]["output_img"].shape[0:2]
@@ -456,7 +460,7 @@ def get_table_recognition_res(
         table_cells_result, crop_start_point, img_shape
     )
 
-    if use_table_cells_ocr_results == True:
+    if use_table_cells_ocr_results == True and use_table_cells_split_ocr == False:
         ocr_dt_boxes = table_cells_result
         ocr_texts_res = cells_texts_list
     else:

paddlex/inference/pipelines/ts_anomaly_detection/pipeline.py

@@ -44,9 +44,9 @@ class TSAnomalyDetPipeline(BasePipeline):
             device (str, optional): Device to run the predictions on. Defaults to None.
             pp_option (PaddlePredictorOption, optional): PaddlePredictor options. Defaults to None.
             use_hpip (bool, optional): Whether to use the high-performance
-                inference plugin (HPIP). Defaults to False.
+                inference plugin (HPIP) by default. Defaults to False.
             hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
-                The high-performance inference configuration dictionary.
+                The default high-performance inference configuration dictionary.
                 Defaults to None.
         """
 

paddlex/inference/pipelines/ts_classification/pipeline.py

@@ -44,9 +44,9 @@ class TSClsPipeline(BasePipeline):
             device (str, optional): Device to run the predictions on. Defaults to None.
             pp_option (PaddlePredictorOption, optional): PaddlePredictor options. Defaults to None.
             use_hpip (bool, optional): Whether to use the high-performance
-                inference plugin (HPIP). Defaults to False.
+                inference plugin (HPIP) by default. Defaults to False.
             hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
-                The high-performance inference configuration dictionary.
+                The default high-performance inference configuration dictionary.
                 Defaults to None.
         """
 

paddlex/inference/pipelines/ts_forecasting/pipeline.py

@@ -44,9 +44,9 @@ class TSFcPipeline(BasePipeline):
             device (str, optional): Device to run the predictions on. Defaults to None.
             pp_option (PaddlePredictorOption, optional): PaddlePredictor options. Defaults to None.
             use_hpip (bool, optional): Whether to use the high-performance
-                inference plugin (HPIP). Defaults to False.
+                inference plugin (HPIP) by default. Defaults to False.
             hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
-                The high-performance inference configuration dictionary.
+                The default high-performance inference configuration dictionary.
                 Defaults to None.
         """
 

paddlex/inference/pipelines/video_classification/pipeline.py

@@ -45,9 +45,9 @@ class VideoClassificationPipeline(BasePipeline):
             device (str): The device to run the prediction on. Default is None.
             pp_option (PaddlePredictorOption): Options for PaddlePaddle predictor. Default is None.
             use_hpip (bool, optional): Whether to use the high-performance
-                inference plugin (HPIP). Defaults to False.
+                inference plugin (HPIP) by default. Defaults to False.
             hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
-                The high-performance inference configuration dictionary.
+                The default high-performance inference configuration dictionary.
                 Defaults to None.
         """
         super().__init__(

paddlex/inference/pipelines/video_detection/pipeline.py

@@ -45,9 +45,9 @@ class VideoDetectionPipeline(BasePipeline):
             device (str): The device to run the prediction on. Default is None.
             pp_option (PaddlePredictorOption): Options for PaddlePaddle predictor. Default is None.
             use_hpip (bool, optional): Whether to use the high-performance
-                inference plugin (HPIP). Defaults to False.
+                inference plugin (HPIP) by default. Defaults to False.
             hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional):
-                The high-performance inference configuration dictionary.
+                The default high-performance inference configuration dictionary.
                 Defaults to None.
         """
         super().__init__(

paddlex/inference/serving/basic_serving/_pipeline_apps/_common/common.py

@@ -90,7 +90,11 @@ def postprocess_images(
     output_images: Dict[str, str] = {}
     for key, img in images.items():
         output_images[key] = postprocess_image(
-            np.array(img) if isinstance(img, Image) else img,
+            (
+                cv2.cvtColor(np.array(img.convert("RGB")), cv2.COLOR_RGB2BGR)
+                if isinstance(img, Image)
+                else img
+            ),
             log_id=log_id,
             filename=filename_template.format(key=key),
             file_storage=file_storage,

paddlex/inference/serving/basic_serving/_pipeline_apps/layout_parsing.py

@@ -54,7 +54,6 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
             use_doc_orientation_classify=request.useDocOrientationClassify,
             use_doc_unwarping=request.useDocUnwarping,
             use_textline_orientation=request.useTextlineOrientation,
-            use_general_ocr=request.useGeneralOcr,
             use_seal_recognition=request.useSealRecognition,
             use_table_recognition=request.useTableRecognition,
             use_formula_recognition=request.useFormulaRecognition,

paddlex/inference/serving/basic_serving/_pipeline_apps/pp_chatocrv3_doc.py

@@ -54,7 +54,6 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
             images,
             use_doc_orientation_classify=request.useDocOrientationClassify,
             use_doc_unwarping=request.useDocUnwarping,
-            use_general_ocr=request.useGeneralOcr,
             use_seal_recognition=request.useSealRecognition,
             use_table_recognition=request.useTableRecognition,
             layout_threshold=request.layoutThreshold,

paddlex/inference/serving/basic_serving/_pipeline_apps/pp_chatocrv4_doc.py

@@ -54,7 +54,7 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
             images,
             use_doc_orientation_classify=request.useDocOrientationClassify,
             use_doc_unwarping=request.useDocUnwarping,
-            use_general_ocr=request.useGeneralOcr,
+            use_textline_orientation=request.useTextlineOrientation,
             use_seal_recognition=request.useSealRecognition,
             use_table_recognition=request.useTableRecognition,
             layout_threshold=request.layoutThreshold,

paddlex/inference/serving/basic_serving/_pipeline_apps/pp_structurev3.py

@@ -54,10 +54,11 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
             use_doc_orientation_classify=request.useDocOrientationClassify,
             use_doc_unwarping=request.useDocUnwarping,
             use_textline_orientation=request.useTextlineOrientation,
-            use_general_ocr=request.useGeneralOcr,
             use_seal_recognition=request.useSealRecognition,
             use_table_recognition=request.useTableRecognition,
             use_formula_recognition=request.useFormulaRecognition,
+            use_chart_recognition=request.useChartRecognition,
+            use_region_detection=request.useRegionDetection,
             layout_threshold=request.layoutThreshold,
             layout_nms=request.layoutNms,
             layout_unclip_ratio=request.layoutUnclipRatio,
@@ -74,7 +75,10 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
             seal_det_box_thresh=request.sealDetBoxThresh,
             seal_det_unclip_ratio=request.sealDetUnclipRatio,
             seal_rec_score_thresh=request.sealRecScoreThresh,
-            use_table_cells_ocr_results=request.useTableCellsOcrResults,
+            use_wired_table_cells_trans_to_html=request.useWiredTableCellsTransToHtml,
+            use_wireless_table_cells_trans_to_html=request.useWirelessTableCellsTransToHtml,
+            use_table_orientation_classify=request.useTableOrientationClassify,
+            use_ocr_results_with_table_cells=request.useOcrResultsWithTableCells,
             use_e2e_wired_table_rec_model=request.useE2eWiredTableRecModel,
             use_e2e_wireless_table_rec_model=request.useE2eWirelessTableRecModel,
         )

paddlex/inference/serving/basic_serving/_pipeline_apps/table_recognition.py

@@ -53,17 +53,13 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
             use_doc_unwarping=request.useDocUnwarping,
             use_layout_detection=request.useLayoutDetection,
             use_ocr_model=request.useOcrModel,
-            layout_threshold=request.layoutThreshold,
-            layout_nms=request.layoutNms,
-            layout_unclip_ratio=request.layoutUnclipRatio,
-            layout_merge_bboxes_mode=request.layoutMergeBboxesMode,
             text_det_limit_side_len=request.textDetLimitSideLen,
             text_det_limit_type=request.textDetLimitType,
             text_det_thresh=request.textDetThresh,
             text_det_box_thresh=request.textDetBoxThresh,
             text_det_unclip_ratio=request.textDetUnclipRatio,
             text_rec_score_thresh=request.textRecScoreThresh,
-            use_table_cells_ocr_results=request.useTableCellsOcrResults,
+            use_ocr_results_with_table_cells=request.useOcrResultsWithTableCells,
         )
 
         table_rec_results: List[Dict[str, Any]] = []

paddlex/inference/serving/basic_serving/_pipeline_apps/table_recognition_v2.py

@@ -53,19 +53,18 @@ def create_pipeline_app(pipeline: Any, app_config: AppConfig) -> "FastAPI":
             use_doc_unwarping=request.useDocUnwarping,
             use_layout_detection=request.useLayoutDetection,
             use_ocr_model=request.useOcrModel,
-            layout_threshold=request.layoutThreshold,
-            layout_nms=request.layoutNms,
-            layout_unclip_ratio=request.layoutUnclipRatio,
-            layout_merge_bboxes_mode=request.layoutMergeBboxesMode,
             text_det_limit_side_len=request.textDetLimitSideLen,
             text_det_limit_type=request.textDetLimitType,
             text_det_thresh=request.textDetThresh,
             text_det_box_thresh=request.textDetBoxThresh,
             text_det_unclip_ratio=request.textDetUnclipRatio,
             text_rec_score_thresh=request.textRecScoreThresh,
-            use_table_cells_ocr_results=request.useTableCellsOcrResults,
             use_e2e_wired_table_rec_model=request.useE2eWiredTableRecModel,
             use_e2e_wireless_table_rec_model=request.useE2eWirelessTableRecModel,
+            use_wired_table_cells_trans_to_html=request.useWiredTableCellsTransToHtml,
+            use_wireless_table_cells_trans_to_html=request.useWirelessTableCellsTransToHtml,
+            use_table_orientation_classify=request.useTableOrientationClassify,
+            use_ocr_results_with_table_cells=request.useOcrResultsWithTableCells,
         )
 
         table_rec_results: List[Dict[str, Any]] = []

paddlex/inference/serving/infra/utils.py

@@ -38,8 +38,8 @@ if is_dep_available("opencv-contrib-python"):
     import cv2
 if is_dep_available("filetype"):
     import filetype
-if is_dep_available("PyMuPDF"):
-    import fitz
+if is_dep_available("pypdfium2"):
+    import pypdfium2 as pdfium
 if is_dep_available("yarl"):
     import yarl
 
@@ -176,31 +176,29 @@ def base64_encode(data: bytes) -> str:
     return base64.b64encode(data).decode("ascii")
 
 
-@function_requires_deps("PyMuPDF", "opencv-contrib-python")
+@function_requires_deps("pypdfium2", "opencv-contrib-python")
 def read_pdf(
     bytes_: bytes, max_num_imgs: Optional[int] = None
 ) -> Tuple[List[np.ndarray], PDFInfo]:
     images: List[np.ndarray] = []
     page_info_list: List[PDFPageInfo] = []
-    with fitz.open("pdf", bytes_) as doc:
-        for page in doc:
-            if max_num_imgs is not None and len(images) >= max_num_imgs:
-                break
-            # TODO: Do not always use zoom=2.0
-            zoom = 2.0
-            deg = 0
-            mat = fitz.Matrix(zoom, zoom).prerotate(deg)
-            pixmap = page.get_pixmap(matrix=mat, alpha=False)
-            image = np.frombuffer(pixmap.samples, dtype=np.uint8).reshape(
-                pixmap.h, pixmap.w, pixmap.n
-            )
-            image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
-            images.append(image)
-            page_info = PDFPageInfo(
-                width=pixmap.w,
-                height=pixmap.h,
-            )
-            page_info_list.append(page_info)
+    doc = pdfium.PdfDocument(bytes_)
+    for page in doc:
+        if max_num_imgs is not None and len(images) >= max_num_imgs:
+            break
+        # TODO: Do not always use zoom=2.0
+        zoom = 2.0
+        deg = 0
+        image = page.render(scale=zoom, rotation=deg).to_pil()
+        image = image.convert("RGB")
+        image = np.array(image)
+        image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
+        images.append(image)
+        page_info = PDFPageInfo(
+            width=image.shape[1],
+            height=image.shape[0],
+        )
+        page_info_list.append(page_info)
     pdf_info = PDFInfo(
         numPages=len(page_info_list),
         pages=page_info_list,

paddlex/inference/serving/schemas/formula_recognition.py

@@ -34,7 +34,7 @@ class InferRequest(ocr.BaseInferRequest):
     useLayoutDetection: Optional[bool] = None
     useDocOrientationClassify: Optional[bool] = None
     useDocUnwarping: Optional[bool] = None
-    layoutThreshold: Optional[float] = None
+    layoutThreshold: Optional[Union[float, dict]] = None
     layoutNms: Optional[bool] = None
     layoutUnclipRatio: Optional[Union[float, Tuple[float, float]]] = None
     layoutMergeBboxesMode: Optional[str] = None

paddlex/inference/serving/schemas/layout_parsing.py

@@ -34,11 +34,10 @@ class InferRequest(ocr.BaseInferRequest):
     useDocOrientationClassify: Optional[bool] = None
     useDocUnwarping: Optional[bool] = None
     useTextlineOrientation: Optional[bool] = None
-    useGeneralOcr: Optional[bool] = None
     useSealRecognition: Optional[bool] = None
     useTableRecognition: Optional[bool] = None
     useFormulaRecognition: Optional[bool] = None
-    layoutThreshold: Optional[float] = None
+    layoutThreshold: Optional[Union[float, dict]] = None
     layoutNms: Optional[bool] = None
     layoutUnclipRatio: Optional[Union[float, Tuple[float, float]]] = None
     layoutMergeBboxesMode: Optional[str] = None

paddlex/inference/serving/schemas/pp_chatocrv3_doc.py

@@ -39,10 +39,9 @@ ANALYZE_IMAGES_ENDPOINT: Final[str] = "/chatocr-visual"
 class AnalyzeImagesRequest(ocr.BaseInferRequest):
     useDocOrientationClassify: Optional[bool] = None
     useDocUnwarping: Optional[bool] = None
-    useGeneralOcr: Optional[bool] = None
     useSealRecognition: Optional[bool] = None
     useTableRecognition: Optional[bool] = None
-    layoutThreshold: Optional[float] = None
+    layoutThreshold: Optional[Union[float, dict]] = None
     layoutNms: Optional[bool] = None
     layoutUnclipRatio: Optional[Union[float, Tuple[float, float], dict]] = None
     layoutMergeBboxesMode: Optional[Union[str, dict]] = None

paddlex/inference/serving/schemas/pp_chatocrv4_doc.py

@@ -42,10 +42,10 @@ ANALYZE_IMAGES_ENDPOINT: Final[str] = "/chatocr-visual"
 class AnalyzeImagesRequest(ocr.BaseInferRequest):
     useDocOrientationClassify: Optional[bool] = None
     useDocUnwarping: Optional[bool] = None
-    useGeneralOcr: Optional[bool] = None
+    useTextlineOrientation: Optional[bool] = None
     useSealRecognition: Optional[bool] = None
     useTableRecognition: Optional[bool] = None
-    layoutThreshold: Optional[float] = None
+    layoutThreshold: Optional[Union[float, dict]] = None
     layoutNms: Optional[bool] = None
     layoutUnclipRatio: Optional[Union[float, Tuple[float, float], dict]] = None
     layoutMergeBboxesMode: Optional[Union[str, dict]] = None

paddlex/inference/serving/schemas/pp_structurev3.py

@@ -32,14 +32,15 @@ INFER_ENDPOINT: Final[str] = "/layout-parsing"
 
 
 class InferRequest(ocr.BaseInferRequest):
-    useDocOrientationClassify: Optional[bool] = None
-    useDocUnwarping: Optional[bool] = None
+    useDocOrientationClassify: Optional[bool] = False
+    useDocUnwarping: Optional[bool] = False
     useTextlineOrientation: Optional[bool] = None
-    useGeneralOcr: Optional[bool] = None
     useSealRecognition: Optional[bool] = None
     useTableRecognition: Optional[bool] = None
     useFormulaRecognition: Optional[bool] = None
-    layoutThreshold: Optional[float] = None
+    useChartRecognition: Optional[bool] = False
+    useRegionDetection: Optional[bool] = None
+    layoutThreshold: Optional[Union[float, dict]] = None
     layoutNms: Optional[bool] = None
     layoutUnclipRatio: Optional[Union[float, Tuple[float, float], dict]] = None
     layoutMergeBboxesMode: Optional[Union[str, dict]] = None
@@ -55,9 +56,12 @@ class InferRequest(ocr.BaseInferRequest):
     sealDetBoxThresh: Optional[float] = None
     sealDetUnclipRatio: Optional[float] = None
     sealRecScoreThresh: Optional[float] = None
-    useTableCellsOcrResults: bool = False
+    useWiredTableCellsTransToHtml: bool = False
+    useWirelessTableCellsTransToHtml: bool = False
+    useTableOrientationClassify: bool = True
+    useOcrResultsWithTableCells: bool = True
     useE2eWiredTableRecModel: bool = False
-    useE2eWirelessTableRecModel: bool = False
+    useE2eWirelessTableRecModel: bool = True
 
 
 class MarkdownData(BaseModel):

paddlex/inference/serving/schemas/seal_recognition.py

@@ -34,7 +34,7 @@ class InferRequest(ocr.BaseInferRequest):
     useDocOrientationClassify: Optional[bool] = None
     useDocUnwarping: Optional[bool] = None
     useLayoutDetection: Optional[bool] = None
-    layoutThreshold: Optional[float] = None
+    layoutThreshold: Optional[Union[float, dict]] = None
     layoutNms: Optional[bool] = None
     layoutUnclipRatio: Optional[Union[float, Tuple[float, float]]] = None
     layoutMergeBboxesMode: Optional[str] = None

paddlex/inference/serving/schemas/table_recognition.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Dict, Final, List, Optional, Tuple, Union
+from typing import Dict, Final, List, Optional
 
 from pydantic import BaseModel
 
@@ -35,17 +35,13 @@ class InferRequest(ocr.BaseInferRequest):
     useDocUnwarping: Optional[bool] = None
     useLayoutDetection: Optional[bool] = None
     useOcrModel: Optional[bool] = None
-    layoutThreshold: Optional[float] = None
-    layoutNms: Optional[bool] = None
-    layoutUnclipRatio: Optional[Union[float, Tuple[float, float]]] = None
-    layoutMergeBboxesMode: Optional[str] = None
     textDetLimitSideLen: Optional[int] = None
     textDetLimitType: Optional[str] = None
     textDetThresh: Optional[float] = None
     textDetBoxThresh: Optional[float] = None
     textDetUnclipRatio: Optional[float] = None
     textRecScoreThresh: Optional[float] = None
-    useTableCellsOcrResults: bool = False
+    useOcrResultsWithTableCells: bool = False
 
 
 class TableRecResult(BaseModel):

paddlex/inference/serving/schemas/table_recognition_v2.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Dict, Final, List, Optional, Tuple, Union
+from typing import Dict, Final, List, Optional
 
 from pydantic import BaseModel
 
@@ -35,19 +35,18 @@ class InferRequest(ocr.BaseInferRequest):
     useDocUnwarping: Optional[bool] = None
     useLayoutDetection: Optional[bool] = None
     useOcrModel: Optional[bool] = None
-    layoutThreshold: Optional[float] = None
-    layoutNms: Optional[bool] = None
-    layoutUnclipRatio: Optional[Union[float, Tuple[float, float], dict]] = None
-    layoutMergeBboxesMode: Optional[Union[str, dict]] = None
     textDetLimitSideLen: Optional[int] = None
     textDetLimitType: Optional[str] = None
     textDetThresh: Optional[float] = None
     textDetBoxThresh: Optional[float] = None
     textDetUnclipRatio: Optional[float] = None
     textRecScoreThresh: Optional[float] = None
-    useTableCellsOcrResults: bool = False
     useE2eWiredTableRecModel: bool = False
     useE2eWirelessTableRecModel: bool = False
+    useWiredTableCellsTransToHtml: bool = False
+    useWirelessTableCellsTransToHtml: bool = False
+    useTableOrientationClassify: bool = True
+    useOcrResultsWithTableCells: bool = True
 
 
 class TableRecResult(BaseModel):

paddlex/inference/utils/hpi.py

@@ -163,8 +163,12 @@ def suggest_inference_backend_and_config(
         # TODO: Is it better to also check the runtime versions of CUDA and
         # cuDNN, and the versions of CUDA and cuDNN used to build `ultra-infer`?
         cuda_version = get_paddle_cuda_version()
+        if not cuda_version:
+            return None, "No CUDA version was found."
         cuda_version = "".join(map(str, cuda_version))
         cudnn_version = get_paddle_cudnn_version()
+        if not cudnn_version:
+            return None, "No cuDNN version was found."
         cudnn_version = "".join(map(str, cudnn_version[:-1]))
         key = f"gpu_cuda{cuda_version}_cudnn{cudnn_version}"
     else:
@@ -231,10 +235,13 @@ def suggest_inference_backend_and_config(
         pseudo_backend = backend_to_pseudo_backend["paddle"]
         assert pseudo_backend in (
             "paddle",
+            "paddle_fp16",
             "paddle_tensorrt",
             "paddle_tensorrt_fp16",
         ), pseudo_backend
-        if pseudo_backend == "paddle_tensorrt":
+        if pseudo_backend == "paddle_fp16":
+            suggested_backend_config.update({"run_mode": "paddle_fp16"})
+        elif pseudo_backend == "paddle_tensorrt":
             suggested_backend_config.update({"run_mode": "trt_fp32"})
         elif pseudo_backend == "paddle_tensorrt_fp16":
             # TODO: Check if the target device supports FP16.

paddlex/inference/utils/hpi_model_info_collection.json

@@ -445,6 +445,14 @@
       "onnxruntime",
       "paddle"
     ],
+    "PP-DocLayout_plus-L": [
+      "onnxruntime",
+      "paddle"
+    ],
+    "PP-DocBlockLayout": [
+      "onnxruntime",
+      "paddle"
+    ],
     "RT-DETR-H_layout_17cls": [
       "onnxruntime",
       "paddle"
@@ -571,9 +579,9 @@
       "paddle"
     ],
     "PP-OCRv4_server_det": [
+      "paddle",
       "openvino",
-      "onnxruntime",
-      "paddle"
+      "onnxruntime"
     ],
     "PP-OCRv3_mobile_rec": [
       "openvino",
@@ -1136,6 +1144,38 @@
     ],
     "YOLO-Worldv2-L": [
       "paddle"
+    ],
+    "PP-OCRv5_server_rec": [
+      "paddle",
+      "openvino",
+      "onnxruntime"
+    ],
+    "PP-OCRv5_mobile_rec": [
+      "openvino",
+      "onnxruntime",
+      "paddle"
+    ],
+    "PP-OCRv5_server_det": [
+      "openvino",
+      "onnxruntime",
+      "paddle"
+    ],
+    "PP-OCRv5_mobile_det": [
+      "openvino",
+      "onnxruntime",
+      "paddle"
+    ],
+    "PP-FormulaNet_plus-L": [
+      "onnxruntime",
+      "paddle"
+    ],
+    "PP-FormulaNet_plus-M": [
+      "onnxruntime",
+      "paddle"
+    ],
+    "PP-FormulaNet_plus-S": [
+      "onnxruntime",
+      "paddle"
     ]
   },
   "gpu_cuda118_cudnn89": {
@@ -2247,6 +2287,45 @@
     ],
     "YOLO-Worldv2-L": [
       "paddle"
+    ],
+    "PP-DocBlockLayout": [
+      "tensorrt",
+      "paddle",
+      "onnxruntime"
+    ],
+    "PP-DocLayout_plus-L": [
+      "tensorrt_fp16",
+      "paddle",
+      "onnxruntime"
+    ],
+    "PP-OCRv5_server_rec": [
+      "paddle_tensorrt_fp16",
+      "tensorrt_fp16",
+      "onnxruntime"
+    ],
+    "PP-OCRv5_mobile_rec": [
+      "paddle_tensorrt_fp16",
+      "tensorrt",
+      "onnxruntime"
+    ],
+    "PP-OCRv5_server_det": [
+      "tensorrt",
+      "onnxruntime",
+      "paddle"
+    ],
+    "PP-OCRv5_mobile_det": [
+      "paddle_tensorrt",
+      "tensorrt",
+      "onnxruntime"
+    ],
+    "PP-FormulaNet_plus-L": [
+      "paddle"
+    ],
+    "PP-FormulaNet_plus-M": [
+      "paddle"
+    ],
+    "PP-FormulaNet_plus-S": [
+      "paddle"
     ]
   }
 }