Update app.py

app.py

@@ -19,16 +19,25 @@ from huggingface_hub import login, HfFolder
 #                                              torch_dtype =torch.float16, 
 #                                              low_cpu_mem_usage=True, 
 #                                              trust_remote_code=True)
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from peft import PeftModel
+# from transformers import AutoModelForCausalLM, AutoTokenizer
+# from peft import PeftModel
 
-# 1. Load Your Base Model and LoRA Adapter
-model_name_or_path = "FlawedLLM/Bhashini_gemma_merged4bit_clean_final"  # Hugging Face model or local path
-lora_weights = "FlawedLLM/Bhashini_gemma_lora_clean_final"  # LoRA weights
+# # 1. Load Your Base Model and LoRA Adapter
+# model_name_or_path = "FlawedLLM/Bhashini_gemma_merged4bit_clean_final"  # Hugging Face model or local path
+# lora_weights = "FlawedLLM/Bhashini_gemma_lora_clean_final"  # LoRA weights
 
-tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
-model = AutoModelForCausalLM.from_pretrained(model_name_or_path, load_in_8bit=True, device_map='auto')
-model = PeftModel.from_pretrained(model, lora_weights)
+# tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
+# model = AutoModelForCausalLM.from_pretrained(model_name_or_path, load_in_8bit=True, device_map='auto')
+# model = PeftModel.from_pretrained(model, lora_weights)
+# Load model directly
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+tokenizer = AutoTokenizer.from_pretrained("FlawedLLM/Bhashini_gemma_merged4bit_clean_final",trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained("FlawedLLM/Bhashini_gemma_merged4bit_clean_final",
+                                             device_map="auto",
+                                             torch_dtype =torch.float16, 
+                                             low_cpu_mem_usage=True, 
+                                             trust_remote_code=True)
 # alpaca_prompt = You MUST copy from above!
 @spaces.GPU(duration=300)
 def chunk_it(input_command, item_list):