fix: set device_map to auto

app.py

@@ -8,11 +8,11 @@ from transformers import (
 import torch
 
 chat_model_name = "sapienzanlp/Minerva-7B-instruct-v1.0"
-chat_model = AutoModelForCausalLM.from_pretrained(chat_model_name, dtype=torch.bfloat16)
+chat_model = AutoModelForCausalLM.from_pretrained(chat_model_name, dtype=torch.bfloat16, device_map="auto")
 chat_tokenizer = AutoTokenizer.from_pretrained(chat_model_name)
 
 moderator_model_name = "saiteki-kai/QA-DeBERTa-v3-large"
-moderator_model = AutoModelForSequenceClassification.from_pretrained(moderator_model_name)
+moderator_model = AutoModelForSequenceClassification.from_pretrained(moderator_model_name, device_map="auto")
 moderator_tokenizer = AutoTokenizer.from_pretrained(moderator_model_name)
 
 @spaces.GPU()