Update app.py

app.py

@@ -7,19 +7,28 @@ import gradio as gr
 import torch
 from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
 from huggingface_hub import login, HfFolder
-tokenizer = AutoTokenizer.from_pretrained("FlawedLLM/Bhashini_gemma_merged16bit_clean_final", trust_remote_code=True)
-quantization_config = BitsAndBytesConfig(
-        load_in_4bit=True,
-        bnb_4bit_use_double_quant=True,
-        bnb_4bit_quant_type="nf4",  
-        bnb_4bit_compute_dtype=torch.float16)
-model = AutoModelForCausalLM.from_pretrained("FlawedLLM/Bhashini_gemma_merged16bit_clean_final", 
-                                             device_map="auto",
-                                             quantization_config=quantization_config, 
-                                             torch_dtype =torch.float16, 
-                                             low_cpu_mem_usage=True, 
-                                             trust_remote_code=True)
+# tokenizer = AutoTokenizer.from_pretrained("FlawedLLM/Bhashini_gemma_merged16bit_clean_final", trust_remote_code=True)
+# quantization_config = BitsAndBytesConfig(
+#         load_in_4bit=True,
+#         bnb_4bit_use_double_quant=True,
+#         bnb_4bit_quant_type="nf4",  
+#         bnb_4bit_compute_dtype=torch.float16)
+# model = AutoModelForCausalLM.from_pretrained("FlawedLLM/Bhashini_gemma_merged16bit_clean_final", 
+#                                              device_map="auto",
+#                                              quantization_config=quantization_config, 
+#                                              torch_dtype =torch.float16, 
+#                                              low_cpu_mem_usage=True, 
+#                                              trust_remote_code=True)
+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
+
+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)
 # alpaca_prompt = You MUST copy from above!
 @spaces.GPU(duration=300)
 def chunk_it(input_command, item_list):