add more log

app.py

@@ -2,26 +2,51 @@ from fastapi import FastAPI, Request
 from pydantic import BaseModel
 from transformers import AutoTokenizer, AutoModel
 import torch
+import time
+import logging
+from datetime import datetime
+
+# Cấu hình logging
+logging.basicConfig(
+    format="%(asctime)s - %(levelname)s - %(message)s",
+    level=logging.INFO
+)
 
 app = FastAPI()
 
 # Load model
 model_name = "AITeamVN/Vietnamese_Embedding_v2"
+logging.info(f"Loading model: {model_name}")
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModel.from_pretrained(model_name)
+logging.info("Model loaded successfully.")
 
 class InputText(BaseModel):
     text: str
 
 @app.get("/")
 def root():
+    now = datetime.now().isoformat()
+    logging.info(f"[GET /] Received health check at {now}")
     return {"message": "AITeamVN/Vietnamese_Embedding_v2 embedding API is running."}
 
 @app.post("/embed")
 def get_embedding(data: InputText):
+    start_time = time.time()
+    start_ts = datetime.now().isoformat()
+
+    # Tokenize input
     inputs = tokenizer(data.text, return_tensors="pt", padding=True, truncation=True)
+    input_token_count = inputs["input_ids"].shape[1]
+    logging.info(f"[POST /embed] Start at {start_ts} | Input text: '{data.text[:50]}'... | Tokens: {input_token_count}")
+
+    # Run model inference
     with torch.no_grad():
         outputs = model(**inputs)
-        # Get CLS token or use pooling method
         embedding = outputs.last_hidden_state[:, 0, :].squeeze().tolist()
+
+    end_ts = datetime.now().isoformat()
+    duration_ms = (time.time() - start_time) * 1000
+    logging.info(f"[POST /embed] Done at {end_ts} | Embedding size: {len(embedding)} | Time: {duration_ms:.2f} ms")
+
     return {"embedding": embedding}