Update app.py

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 import spaces
 import torch
+from torch.cuda.amp import autocast
 import subprocess
 from huggingface_hub import InferenceClient
 import os
@@ -46,7 +47,7 @@ with init_empty_weights():
                                              attn_implementation="flash_attention_2",
                                              # low_cpu_mem_usage=True,
                                              # llm_int8_enable_fp32_cpu_offload=True,
-                                             device_map="cuda",
+                                             # device_map="cuda",
                                              
                                             )
 
@@ -57,7 +58,9 @@ with init_empty_weights():
 device_map = infer_auto_device_map(model, max_memory={0: "80GB", 1: "80GB", "cpu": "65GB"})
 
 # Load the model with the inferred device map
-model = load_checkpoint_and_dispatch(model, "path_to_checkpoint", device_map=device_map, no_split_module_classes=["GPTJBlock"])
+model = load_checkpoint_and_dispatch(model, "path_to_checkpoint", device_map=device_map)
+model.half()
+
 
 
 @spaces.GPU(duration=60)