update app

app.py

@@ -3,7 +3,7 @@ import yfinance as yf
 import pandas as pd
 import numpy as np
 import torch
-from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, T5Tokenizer
+from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
 from datetime import datetime, timedelta
 import plotly.graph_objects as go
 import plotly.express as px
@@ -30,19 +30,18 @@ from config import IDX_STOCKS, TECHNICAL_INDICATORS, PREDICTION_CONFIG
 @spaces.GPU(duration=120)
 def load_model():
     """Load the Amazon Chronos-Bolt model for time series forecasting"""
-    # FIX 1: Use AutoModelForSeq2SeqLM and trust_remote_code=True (correct for T5-based model)
+    # Fix: Use AutoModelForSeq2SeqLM and trust_remote_code=True for the T5-based model
     model = AutoModelForSeq2SeqLM.from_pretrained(
         "amazon/chronos-bolt-base",
         torch_dtype=torch.bfloat16,
         device_map="auto",
         trust_remote_code=True
     )
-    # FIX 2: Explicitly use T5Tokenizer (slow version) with trust_remote_code=True and use_fast=False
-    # This bypasses the faulty AutoTokenizer/fast conversion logic.
-    tokenizer = T5Tokenizer.from_pretrained(
+    # Fix: Reverting to AutoTokenizer with ONLY trust_remote_code=True 
+    # This should correctly invoke the custom ChronosTokenizer class and resolve model files.
+    tokenizer = AutoTokenizer.from_pretrained(
         "amazon/chronos-bolt-base", 
-        trust_remote_code=True,
-        use_fast=False
+        trust_remote_code=True
     )
     return model, tokenizer