Update app.py

app.py

@@ -5,11 +5,11 @@ import subprocess
 import sys
 
 # Force install the specific transformers version from the GitHub PR
-subprocess.check_call([sys.executable, "-m", "pip", "install", "--force-reinstall", "--no-deps", "git+https://github.com/Muennighoff/transformers.git@olmoe"])
+subprocess.check_call([sys.executable, "-m", "pip", "install", "-U", "--force-reinstall", "accelerate", "git+https://github.com/Muennighoff/transformers.git@olmoe"])
 
 from transformers import OlmoeForCausalLM, AutoTokenizer
 
-model_name = "allenai/OLMoE-1B-7B-0924-Instruct"
+model_name = "allenai/OLMoE-1B-7B-0924"
 
 # Wrap model loading in a try-except block to handle potential errors
 try:
@@ -17,7 +17,7 @@ try:
     model = OlmoeForCausalLM.from_pretrained(
         model_name, 
         trust_remote_code=True, 
-        torch_dtype=torch.float16 if DEVICE == "cuda" else torch.float32,
+        torch_dtype=torch.bfloat16 if DEVICE == "cuda" else torch.float32,
         low_cpu_mem_usage=True,
         device_map="auto"
     )
@@ -55,7 +55,7 @@ def generate_response(message, history, temperature, max_new_tokens):
 
 css = """
   #output {
-    height: 500px; 
+    height: 900px; 
     overflow: auto; 
     border: 1px solid #ccc; 
   }
@@ -85,5 +85,5 @@ with gr.Blocks(css=css) as demo:
     clear.click(lambda: None, None, chatbot, queue=False)
 
 if __name__ == "__main__":
-    demo.queue(api_open=False)
-    demo.launch(debug=True, show_api=False)
+    demo.queue(api_open=True)
+    demo.launch(debug=True, show_api=True, share=True)