Update app.py

app.py

@@ -8,9 +8,14 @@ models=[
     "google/gemma-7b",
     "google/gemma-7b-it",
     "google/gemma-2b",
-    "google/gemma-2b-it",
+    "google/gemma-2b-it"
+]
+clients=[
+InferenceClient(models[0]),
+InferenceClient(models[1]),
+InferenceClient(models[2]),
+InferenceClient(models[3]),
 ]
-client_z=[]
 
 VERBOSE=False
 
@@ -37,8 +42,9 @@ def format_prompt(message, history, cust_p):
 
 def chat_inf(system_prompt,prompt,history,memory,client_choice,seed,temp,tokens,top_p,rep_p,chat_mem,cust_p):
     #token max=8192
+    print(client_choice)
     hist_len=0
-    client=client_z[0]
+    client=clients[int(client_choice)-1]
     if not history:
         history = []
         hist_len=0