let's load model from local (#163)

deepdoc/parser/pdf_parser.py

 from deepdoc.vision import OCR, Recognizer, LayoutRecognizer, TableStructureRecognizer
 from rag.nlp import huqie
 from copy import deepcopy
-from huggingface_hub import hf_hub_download, snapshot_download
+from huggingface_hub import snapshot_download
 
 logging.getLogger("pdfminer").setLevel(logging.WARNING)
 
         if torch.cuda.is_available():
             self.updown_cnt_mdl.set_param({"device": "cuda"})
         try:
-            model_dir = snapshot_download(
-                repo_id="InfiniFlow/text_concat_xgb_v1.0",
-                local_dir=os.path.join(
+            model_dir = os.path.join(
                     get_project_base_directory(),
-                    "rag/res/deepdoc"),
-                local_files_only=True)
+                    "rag/res/deepdoc")
+            self.updown_cnt_mdl.load_model(os.path.join(
+                model_dir, "updown_concat_xgb.model"))
         except Exception as e:
             model_dir = snapshot_download(
                 repo_id="InfiniFlow/text_concat_xgb_v1.0")
+            self.updown_cnt_mdl.load_model(os.path.join(
+                model_dir, "updown_concat_xgb.model"))
+
 
-        self.updown_cnt_mdl.load_model(os.path.join(
-            model_dir, "updown_concat_xgb.model"))
         self.page_from = 0
         """
         If you have trouble downloading HuggingFace models, -_^ this might help!!

deepdoc/vision/layout_recognizer.py

 import numpy as np
 from huggingface_hub import snapshot_download
 
-from api.db import ParserType
 from api.utils.file_utils import get_project_base_directory
 from deepdoc.vision import Recognizer
 
 
     def __init__(self, domain):
         try:
-            model_dir = snapshot_download(
-                repo_id="InfiniFlow/deepdoc",
-                local_dir=os.path.join(
+            model_dir = os.path.join(
                     get_project_base_directory(),
-                    "rag/res/deepdoc"),
-                local_files_only=True)
+                    "rag/res/deepdoc")
+            super().__init__(self.labels, domain, model_dir)
         except Exception as e:
             model_dir = snapshot_download(repo_id="InfiniFlow/deepdoc")
+            super().__init__(self.labels, domain, model_dir)
 
-        # os.path.join(get_project_base_directory(), "rag/res/deepdoc/"))
-        super().__init__(self.labels, domain, model_dir)
         self.garbage_layouts = ["footer", "header", "reference"]
 
     def __call__(self, image_list, ocr_res, scale_factor=3,

deepdoc/vision/ocr.py

         """
         if not model_dir:
             try:
-                model_dir = snapshot_download(
-                    repo_id="InfiniFlow/deepdoc",
-                    local_dir=os.path.join(
+                model_dir = os.path.join(
                         get_project_base_directory(),
-                        "rag/res/deepdoc"),
-                    local_files_only=True)
+                        "rag/res/deepdoc")
+                self.text_detector = TextDetector(model_dir)
+                self.text_recognizer = TextRecognizer(model_dir)
             except Exception as e:
                 model_dir = snapshot_download(repo_id="InfiniFlow/deepdoc")
+                self.text_detector = TextDetector(model_dir)
+                self.text_recognizer = TextRecognizer(model_dir)
 
-        self.text_detector = TextDetector(model_dir)
-        self.text_recognizer = TextRecognizer(model_dir)
         self.drop_score = 0.5
         self.crop_image_res_index = 0
 

deepdoc/vision/recognizer.py

 
         """
         if not model_dir:
-            try:
-                model_dir = snapshot_download(
-                    repo_id="InfiniFlow/deepdoc",
-                    local_dir=os.path.join(
+            model_dir = os.path.join(
                         get_project_base_directory(),
-                        "rag/res/deepdoc"),
-                    local_files_only=True)
-            except Exception as e:
+                        "rag/res/deepdoc")
+            model_file_path = os.path.join(model_dir, task_name + ".onnx")
+            if not os.path.exists(model_file_path):
                 model_dir = snapshot_download(repo_id="InfiniFlow/deepdoc")
+                model_file_path = os.path.join(model_dir, task_name + ".onnx")
 
-        model_file_path = os.path.join(model_dir, task_name + ".onnx")
         if not os.path.exists(model_file_path):
             raise ValueError("not find model file path {}".format(
                 model_file_path))

deepdoc/vision/table_structure_recognizer.py

 
     def __init__(self):
         try:
-            model_dir = snapshot_download(
-                repo_id="InfiniFlow/deepdoc",
-                local_dir=os.path.join(
+            super().__init__(self.labels, "tsr", os.path.join(
                     get_project_base_directory(),
-                    "rag/res/deepdoc"),
-                local_files_only=True)
+                    "rag/res/deepdoc"))
         except Exception as e:
-            model_dir = snapshot_download(repo_id="InfiniFlow/deepdoc")
-
-        # os.path.join(get_project_base_directory(), "rag/res/deepdoc/"))
-        super().__init__(self.labels, "tsr", model_dir)
+            super().__init__(self.labels, "tsr", snapshot_download(repo_id="InfiniFlow/deepdoc"))
 
     def __call__(self, images, thr=0.2):
         tbls = super().__call__(images, thr)

rag/llm/embedding_model.py

 from rag.utils import num_tokens_from_string
 
 try:
-    model_dir = snapshot_download(
-        repo_id="BAAI/bge-large-zh-v1.5",
-        local_dir=os.path.join(
+    flag_model = FlagModel(os.path.join(
             get_project_base_directory(),
             "rag/res/bge-large-zh-v1.5"),
-        local_files_only=True)
+                           query_instruction_for_retrieval="为这个句子生成表示以用于检索相关文章：",
+                           use_fp16=torch.cuda.is_available())
 except Exception as e:
-    model_dir = snapshot_download(repo_id="BAAI/bge-large-zh-v1.5")
-
-flag_model = FlagModel(model_dir,
+    flag_model = FlagModel("BAAI/bge-large-zh-v1.5",
                        query_instruction_for_retrieval="为这个句子生成表示以用于检索相关文章：",
                        use_fp16=torch.cuda.is_available())
 

rag/nlp/search.py

                       for ck in chunks]
         cites = {}
         thr = 0.63
-        while len(cites.keys()) == 0 and pieces_ and chunks_tks:
+        while thr>0.3 and len(cites.keys()) == 0 and pieces_ and chunks_tks:
             for i, a in enumerate(pieces_):
                 sim, tksim, vtsim = self.qryr.hybrid_similarity(ans_v[i],
                                                                 chunk_v,