Update app.py

app.py

@@ -21,6 +21,9 @@ subprocess.run(
     shell=True,
 )
 
+import bitsandbytes as bnb  # Import bitsandbytes for 8-bit quantization
+
+
 subprocess.run(
     "pip install pynvml gpustat",
    
@@ -74,11 +77,12 @@ accelerator = Accelerator()
 
 model = AutoModelForCausalLM.from_pretrained(model_id, token= token, 
                                                  # torch_dtype= torch.uint8, 
-                                             torch_dtype=torch.float16,
+                                             # torch_dtype=torch.float16,
+                                              # load_in_8bit=True,
                                               # torch_dtype=torch.fl,
-                                             attn_implementation="flash_attention_2",
-                                             low_cpu_mem_usage=True,
-                                            
+                                             # attn_implementation="flash_attention_2",
+                                             # low_cpu_mem_usage=True,
+                                            # quantization_config=bnb.QuantizationConfig(bits=8), 
                                              # device_map='cuda',
                                              # device_map=accelerator.device_map,