Update app.py

app.py

@@ -5,8 +5,7 @@ from huggingface_hub import InferenceClient
 Copied from inference in colab notebook
 """
 
-from transformers import AutoTokenizer , AutoModelForCausalLM , TextIteratorStreamer
-import torch
+from transformers import AutoTokenizer , AutoModelForSeq2SeqLM , TextIteratorStreamer
 from threading import Thread
 
 # Load model and tokenizer globally to avoid reloading for every request
@@ -17,7 +16,7 @@ model_path = "Mat17892/t5small_enfr_opus"
 tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=True, legacy=False)
 
 # Load the base model (e.g., LLaMA)
-base_model = AutoModelForCausalLM.from_pretrained(base_model)
+base_model = AutoModelForSeq2SeqLM.from_pretrained(base_model)
 
 # Load LoRA adapter
 from peft import PeftModel