fix: add cache folder for transformers

app/api/endpoints/analysis.py

@@ -2,18 +2,15 @@ from typing import Any, List, Dict
 import uuid
 from datetime import datetime
 from fastapi import APIRouter, HTTPException, status, Request
+from trendspy import Trends
 
 import asyncio
 from motor.motor_asyncio import AsyncIOMotorClient
 from bson import ObjectId
 
-import pandas as pd
-from trendspy import Trends
-
 from app.core.config import settings
 from app.core.clients import qstash_client
 from app.schemas.analysis_schema import (
-    WeeklyTrendResponseSchema,
     WeeklyTrendListResponse,
     TrendDetailResponseSchema,
     OnDemandRequestSchema,

app/core/config.py

@@ -34,6 +34,10 @@ class Settings(BaseSettings):
     CONSUMER_BATCH_SIZE: int = 32
     CONSUMER_BATCH_TIMEOUT_SECONDS: int = 5
 
+    # Cache folder
+    HF_HOME: str = "/data"
+    TRANSFORMERS_CACHE: str = "/data/transformers"
+
     # FastAPI
     API_PREFIX: str = "/api"
     API_VERSION: str = "/v1"

app/services/sentiment_service.py

@@ -27,11 +27,13 @@ class SentimentService:
 
         # Load model, tokenizer, and config (for id2label mapping)
         model_name = settings.SENTIMENT_MODEL
-        self.tokenizer = AutoTokenizer.from_pretrained(model_name)
-        self.config = AutoConfig.from_pretrained(model_name)
-        self.model = AutoModelForSequenceClassification.from_pretrained(model_name).to(
-            self.device
+        self.tokenizer = AutoTokenizer.from_pretrained(
+            model_name, cache_dir=settings.TRANSFORMERS_CACHE
         )
+        self.config = AutoConfig.from_pretrained(model_name)
+        self.model = AutoModelForSequenceClassification.from_pretrained(
+            model_name, cache_dir=settings.TRANSFORMERS_CACHE
+        ).to(self.device)
         self.model.eval()  # set model to inference mode
         print("Sentiment model loaded successfully.")