Update app.py

app.py

@@ -152,7 +152,7 @@ def reset_chat(idx, ld, state, palm_if):
 
 async def chat_stream(
     idx, local_data, instruction_txtbox, chat_state,
-    global_context, res_temp, res_topk, res_rpen, res_mnts, res_sample, ctx_num_lconv,
+    global_context, res_temp, res_topk, res_topp, 
     internet_option, serper_api_key, palm_if
 ):
     internet_option = True if internet_option == "on" else False
@@ -171,9 +171,17 @@ async def chat_stream(
     #         search_prompt = tmp_prompt
     #         yield "", prompt, uis, str(res), gr.update(interactive=False), "off"
 
+    parameters = {
+        'model': 'models/text-bison-001',
+        'candidate_count': 1,
+        'temperature': res_temp,
+        'top_k': res_topk,
+        'top_p': res_topp,
+    }
+    
     palm_if, response_txt = gen_text(
         instruction_txtbox, palm_if,
-        global_context, res_temp, res_topk
+        parameters=parameters
     )
 
     ppm = res[idx]
@@ -338,15 +346,8 @@ with gr.Blocks(css=MODEL_SELECTION_CSS, theme='gradio/soft') as demo:
                     with gr.Row():
                         res_temp = gr.Slider(0.0, 2.0, 1.0, step=0.1, label="temp", interactive=True)
                         res_topk = gr.Slider(20, 1000, 50, step=1, label="top_k", interactive=True)
-                        res_rpen = gr.Slider(0.0, 2.0, 1.2, step=0.1, label="rep_penalty", interactive=True)
-                        res_mnts = gr.Slider(64, 8192, 512, step=1, label="new_tokens", interactive=True)
-                        res_sample = gr.Radio([True, False], value=True, label="sample", interactive=True)
+                        res_topp = gr.Slider(0.0, 1.0, 0.9, step=0.1, label="top_p", interactive=True)
                 
-                with gr.Column():
-                    gr.Markdown("#### Context managements")
-                    with gr.Row():
-                        ctx_num_lconv = gr.Slider(2, 10, 3, step=1, label="number of recent talks to keep", interactive=True)
-
         gr.Markdown(
             "***NOTE:*** If you are subscribing [PRO](https://huggingface.co/pricing#pro), you can simply duplicate this space and use your "
             "Hugging Face Access Token to run the same application. Just add `HF_TOKEN` secret with the Token value accorindg to [this guide]"
@@ -370,7 +371,7 @@ with gr.Blocks(css=MODEL_SELECTION_CSS, theme='gradio/soft') as demo:
     ).then(
         chat_stream,
         [idx, local_data, instruction_txtbox, chat_state,
-         global_context, res_temp, res_topk, res_rpen, res_mnts, res_sample, ctx_num_lconv,
+         global_context, res_temp, res_topk, res_topp,
          internet_option, serper_api_key, palm_if],
         [instruction_txtbox, context_inspector, chatbot, local_data, regenerate, internet_option]
     ).then(
@@ -402,7 +403,7 @@ with gr.Blocks(css=MODEL_SELECTION_CSS, theme='gradio/soft') as demo:
     regen_event = regenerate.click(
         rollback_last,
         [idx, local_data, chat_state,
-         global_context, res_temp, res_topk, res_rpen, res_mnts, res_sample, ctx_num_lconv,
+         global_context, res_temp, res_topk, res_topp
          internet_option, serper_api_key, palm_if],
         [context_inspector, chatbot, local_data, regenerate, internet_option]
     ).then(