Update app.py

app.py

@@ -9,42 +9,48 @@ import json
 print("Version ---- 1")
 app = FastAPI()
 
-from huggingface_hub import snapshot_download
+from huggingface_hub import snapshot_download, hf_hub_download
+import os
+import shutil
 
-# Path expected by sinatools
 BASE_DIR = os.path.expanduser("~/.sinatools")
-MODEL_DIR = os.path.join(BASE_DIR, "relation_model")
 
-# Create directory
+# Paths expected by sinatools
+RELATION_MODEL_DIR = os.path.join(BASE_DIR, "relation_model")
+NER_DIR = os.path.join(BASE_DIR, "Wj27012000.tar")
+
 os.makedirs(BASE_DIR, exist_ok=True)
 
-# Download model ONLY if not exists
-if not os.path.exists(MODEL_DIR) or not os.listdir(MODEL_DIR):
+# -------------------------
+# 1. Download relation model
+# -------------------------
+if not os.path.exists(RELATION_MODEL_DIR) or not os.listdir(RELATION_MODEL_DIR):
     snapshot_download(
         repo_id="aaljabari/arabic-relation-extraction-model",
-        local_dir=MODEL_DIR,
+        local_dir=RELATION_MODEL_DIR,
         local_dir_use_symlinks=False
     )
 
+# -------------------------
+# 2. Download NER resources
+# -------------------------
+if not os.path.exists(NER_DIR):
+    os.makedirs(NER_DIR, exist_ok=True)
 
-pretrained_path = "aubmindlab/bert-base-arabertv2"  # must match training
-tokenizer = AutoTokenizer.from_pretrained(pretrained_path)
-encoder = AutoModel.from_pretrained(pretrained_path).eval()
-
+    nested_repo_path = snapshot_download(
+        repo_id="SinaLab/Nested"
+    )
 
-checkpoint_path = snapshot_download(repo_id="SinaLab/Nested", allow_patterns="checkpoints/")
+    # Copy tag_vocab.pkl to expected location
+    src_vocab = os.path.join(nested_repo_path, "Nested", "utils", "tag_vocab.pkl")
+    dst_vocab = os.path.join(NER_DIR, "tag_vocab.pkl")
 
-args_path = hf_hub_download(
-    repo_id="SinaLab/Nested",
-    filename="args.json"
-)
+    if os.path.exists(src_vocab):
+        shutil.copy(src_vocab, dst_vocab)
 
-with open(args_path, 'r') as f:
-    args_data = json.load(f)
-    
-# Load model
-with open("Nested/utils/tag_vocab.pkl", "rb") as f:
-    label_vocab = pickle.load(f)
+# Optional debug
+print("sinatools dir:", os.listdir(BASE_DIR))
+print("NER dir:", os.listdir(NER_DIR))
 
 
 from sinatools.relations.relation_extractor import relation_extraction