Update app.py

app.py

@@ -13,20 +13,10 @@ text4 = "<h1 style='text-align: center; color: blue; font-size: 25px;'>Results"
 diy_value = 0
 saas_value = 0
 
-def calculate_tco(model_choice, vm_rental_choice, out_diy):
-    VM_cost_per_hour=3.6730 #at Azure for the basic pay as you go option
-    maxed_out = 0.8 #percentage of time the VM is maxed out
-    used = 0.5 #percentage of time the VM is used 
+def calculate_tco(maxed_out, used, tokens_per_second_inp, vm_cost_per_hour, vm_rental_choice, out_diy):
     tokens_per_request = 64
-    
-    if model_choice == "Llama-2-7B":
-        tokens_per_second=694.38
-        
-    elif model_choice == "Llama-2-13B":
-        tokens_per_second=1000
-    
-    elif model_choice == "Llama-2-70B":
-        tokens_per_second=10000
+    maxed_out = maxed_out / 100
+    used = used / 100
     
     if vm_rental_choice == "pay as you go":
         reduction = 0
@@ -37,7 +27,7 @@ def calculate_tco(model_choice, vm_rental_choice, out_diy):
     elif vm_rental_choice == "3 years reserved":
         reduction = 0.62
         
-    homemade_cost_per_token = VM_cost_per_hour * (1 - reduction) / (tokens_per_second * 3600 * maxed_out * used)
+    homemade_cost_per_token = vm_cost_per_hour * (1 - reduction) / (tokens_per_second_inp * 3600 * maxed_out * used)
     homemade_cost_per_request = tokens_per_request * homemade_cost_per_token
     out_diy = homemade_cost_per_token
     return out_diy
@@ -55,6 +45,16 @@ def calculate_tco_2(model_provider, context, out_saas):
     out_saas = saas_cost_per_token
     return out_saas
 
+def update_tco(maxed_out, used, tokens_per_second_inp, vm_cost_per_hour_inp, rental_plan_inp, out_diy):
+    if maxed_out!=None and used!=None and tokens_per_second_inp!=None and vm_cost_per_hour_inp!=None and rental_plan_inp!=None:
+        return calculate_tco(maxed_out, used, tokens_per_second_inp, vm_cost_per_hour_inp, rental_plan_inp, out_diy)
+    return None
+
+def update_tco2(model_provider_inp, context_inp, out_saas):
+    if model_provider_inp!=None and context_inp!=None:
+        return calculate_tco_2(model_provider_inp, context_inp, out_saas)
+    return None
+
 def extract_cost_from_text(text):
     try:
         cost = float(text)
@@ -83,13 +83,38 @@ def compare(cost_text1, cost_text2):
         return f"Error: {str(e)}"
 
 def update_plot(diy_value, saas_value):
-                    data = pd.DataFrame(
+    # if maxed_out and used and tokens_per_second_inp and vm_cost_per_hour:
+    #     diy_value = calculate_tco(maxed_out.value, used.value, tokens_per_second_inp, vm_cost_per_hour, vm_rental_choice, out_diy)
+    # else :
+    #     diy_value = 0
+    # if model_provider_inp2 and context_inp2:
+    #     saas_value = calculate_tco_2(model_provider_inp2, context_inp2, out_saas2)
+    # else: 
+    #     saas_value = 0
+    data = pd.DataFrame(
                         {
                             "Solution": ["Open-source", "SaaS"],
                             "Cost/token ($)": [diy_value, saas_value],
                         }
                     )
-                    return gr.BarPlot.update(data, x="Solution", y="Cost/token ($)")
+    return gr.BarPlot.update(data, x="Solution", y="Cost/token ($)")
+
+def update_plot2(diy_value, saas_value):
+    # if maxed_out2!=None and used2!=None and tokens_per_second_inp2!=None and vm_cost_per_hour2!=None and vm_rental_choice!=None:
+    #     diy_value = calculate_tco(maxed_out2.value, used2.value, tokens_per_second_inp2, vm_cost_per_hour2, vm_rental_choice, out_diy2)
+    # else: 
+    #     diy_value = 0
+    # if model_provider_inp2 and context_inp2:
+    #     saas_value = calculate_tco_2(model_provider_inp, context_inp, out_saas)
+    # else: 
+    #     saas_value = 0
+    data = pd.DataFrame(
+                        {
+                            "Solution": ["Open-source", "SaaS"],
+                            "Cost/token ($)": [diy_value, saas_value],
+                        }
+                    )
+    return gr.BarPlot.update(data, x="Solution", y="Cost/token ($)")
 
 def render_latex(latex_str):
     fig, ax = plt.subplots(figsize=(1, 1))
@@ -103,6 +128,21 @@ def render_latex(latex_str):
     base64_str = base64.b64encode(buf.getvalue()).decode("utf-8")
     return f"<img src='data:image/png;base64,{base64_str}'>"
 
+def update_vm_choice(model_inp):
+    if model_inp == "Llama-2-7B" or "Llama-2-13B" or "Llama-2-70B":
+        new_options = ["A100 40GB"]
+        return gr.Dropdown.update(choices=new_options)
+
+def token_per_s_and_cost(vm_inp):
+    if vm_inp == "A100 40GB":
+        return [694.38, 3.6730, 694.38, 3.6730]
+
+def submit_diy(rental_plan):
+    calculate_tco(maxed_out.value, used.value, tokens_per_second_inp.value, vm_cost_per_hour_inp.value, rental_plan, out_diy)
+
+def submit_saas(context_inp):
+    calculate_tco_2(model_provider_inp, context_inp, out_saas)
+
 description=f"""
 <p>In this demo application, we help you compare different solutions for your AI incorporation plans, such as open-source or SaaS.</p>
 <p>First, you'll have to choose the two solutions you'd like to compare. Then, follow the instructions to select your configurations for each solution and we will compute the cost/request accordingly to them. Eventually, you can compare both solutions to evaluate which one best suits your needs, in the short or long term.</p>
@@ -111,8 +151,10 @@ description1="This interface provides you with the cost per token you get using
 description2="This interface provides you with the cost per token resulting from the AI model provider you choose and the number of tokens you select for context, which the model will take into account when processing input texts."
 description3="This interface compares the cost per request for the two solutions you selected and gives you an insight of whether a solution is more valuable in the long term."
 
+test_list = []
 models = ["Llama-2-7B", "Llama-2-13B", "Llama-2-70B"]
 vm_rental_choice = ["pay as you go", "1 year reserved", "3 years reserved"]
+vm_choice = ["A100 40GB"]
 model_provider = ["OpenAI"]
 context = ["4K context", "16K context"]
 error_box = gr.Textbox(label="Error", visible=False)
@@ -125,11 +167,15 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
     out_saas = gr.State(value=0)
     out_diy2 = gr.State(value=0)
     out_saas2 = gr.State(value=0)
+    tokens_per_second_inp = gr.State()
+    vm_cost_per_hour_inp = gr.State()
+    tokens_per_second_inp2 = gr.State()
+    vm_cost_per_hour_inp2 = gr.State()
     
     with gr.Row():
         with gr.Column():
         
-            solution_selection = gr.Dropdown(["SaaS", "Open-source"], label="Select a Solution", value="SaaS")
+            solution_selection = gr.Dropdown(["SaaS", "Open-source"], label="Select a Solution")
             
             with gr.Row(visible=False) as title_column:            
                 gr.Markdown(value=text1)
@@ -143,10 +189,33 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
                 )
             
             with gr.Row(visible=False) as input_diy_column:
-                model_inp = gr.Dropdown(models, label="Select an AI Model", value="Llama-2-7B", info="Open-source AI model used for your application")
-                rental_plan_inp = gr.Dropdown(vm_rental_choice, label="Select a VM Rental Plan", value="pay as you go", info="These options are from Azure's VM rental plans")
-                rental_plan_inp.change(fn=calculate_tco, inputs=[model_inp, rental_plan_inp, out_diy], outputs=out_diy)
-
+                with gr.Column():
+                    
+                    with gr.Row():
+                        model_inp = gr.Dropdown(models, label="Select an AI Model", info="Open-source AI model used for your application")
+                        
+                    with gr.Row() as vm:
+                        with gr.Column():
+                            with gr.Row():
+                                vm_inp = gr.Dropdown(vm_choice, label="Select a Virtual Machine", info="Your options for this choice depend on the model you previously chose")
+                            with gr.Row(visible=False) as vm_info:
+                                token_per_seconds = gr.Textbox(interactive=False, label="Token/s", info="To compute this value based on your model and VM choice, we chose an input length of 233 tokens.")
+                                vm_cost_per_hour = gr.Textbox(interactive=False, label="Cost/h ($) for the VM")                                                               
+                    
+                    with gr.Row() as use_case:
+                        maxed_out = gr.Slider(minimum=0.01, value=80, label="% maxed out", info="percentage of how much your machine is maxed out")
+                        used = gr.Slider(minimum=0.01, value=50, label="% used", info="percentage of time your machine is used")
+                        rental_plan_inp = gr.Dropdown(vm_rental_choice, label="Select a VM Rental Plan", info="These options are from Azure's VM rental plans. By default, the cost taken into account are from the pay as you go plan.")
+                    
+                    model_inp.change(fn=update_vm_choice, inputs=model_inp, outputs=vm_inp)
+                    vm_inp.change(fn=token_per_s_and_cost, inputs=vm_inp, outputs=[tokens_per_second_inp, vm_cost_per_hour_inp, token_per_seconds, vm_cost_per_hour])
+                    
+                    maxed_out.change(fn=update_tco, inputs=[maxed_out, used, tokens_per_second_inp, vm_cost_per_hour_inp, rental_plan_inp, out_diy], outputs=out_diy)
+                    used.change(fn=update_tco, inputs=[maxed_out, used, tokens_per_second_inp, vm_cost_per_hour_inp, rental_plan_inp, out_diy], outputs=out_diy)
+                    model_inp.change(fn=update_tco, inputs=[maxed_out, used, tokens_per_second_inp, vm_cost_per_hour_inp, rental_plan_inp, out_diy], outputs=out_diy)
+                    vm_inp.change(fn=update_tco, inputs=[maxed_out, used, tokens_per_second_inp, vm_cost_per_hour_inp, rental_plan_inp, out_diy], outputs=out_diy)
+                    rental_plan_inp.change(fn=update_tco, inputs=[maxed_out, used, tokens_per_second_inp, vm_cost_per_hour_inp, rental_plan_inp, out_diy], outputs=out_diy)
+                    
             with gr.Row(visible=False) as text_saas_column:
                 gr.Markdown(description2)
             
@@ -156,10 +225,19 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
                 )
             
             with gr.Row(visible=False) as input_saas_column:
-                model_provider_inp = gr.Dropdown(model_provider, label="Model Provider", vallue="OpenAI", info="Choose an AI model provider you want to work with")
-                context_inp = gr.Dropdown(context, label="Context", value="4K context", info="Number of tokens the model considers when processing text")
-                context_inp.change(fn=calculate_tco_2, inputs=[model_provider_inp, context_inp, out_saas], outputs=out_saas)
-    
+                model_provider_inp = gr.Dropdown(model_provider, label="Model Provider", value="OpenAI", info="Choose an AI model provider you want to work with")
+                context_inp = gr.Dropdown(context, label="Context", info="Number of tokens the model considers when processing text")
+                
+                model_provider_inp.change(fn=update_tco2, inputs=[model_provider_inp, context_inp, out_saas], outputs=out_saas)
+                context_inp.change(fn=update_tco2, inputs=[model_provider_inp, context_inp, out_saas], outputs=out_saas)
+
+            def show_vm_info():
+                            return {
+                                vm_info: gr.update(visible=True),
+                            }
+
+            vm_inp.change(show_vm_info, outputs=vm_info)
+            
             def submit(solution_selection):
                 if solution_selection == "Open-source":
                     return {
@@ -185,14 +263,14 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
             solution_selection.change(
                 submit,
                 solution_selection,
-                [out_saas, text_diy_column, formula_diy, formula_saas, title_column, text_saas_column, model_inp, rental_plan_inp, model_provider_inp, context_inp, input_diy_column, input_saas_column],
+                [model_inp, vm, vm_info, vm_inp, maxed_out, used, out_saas, text_diy_column, formula_diy, formula_saas, title_column, text_saas_column, model_inp, rental_plan_inp, model_provider_inp, context_inp, input_diy_column, input_saas_column],
             )
             
         # gr.Divider(style="vertical", thickness=2, color="blue")
          
         with gr.Column():
             
-            solution_selection2 = gr.Dropdown(["SaaS", "Open-source"], value="Open-source", label="Select a Solution")
+            solution_selection2 = gr.Dropdown(["SaaS", "Open-source"], label="Select a solution")
             
             with gr.Row(visible=False) as title_column2:            
                 gr.Markdown(value=text2)
@@ -206,9 +284,31 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
                 )
             
             with gr.Row(visible=False) as input_diy_column2:
-                model_inp2 = gr.Dropdown(models, label="Select an AI Model", value="Llama-2-7B", info="Open-source AI model used for your application")
-                rental_plan_inp2 = gr.Dropdown(vm_rental_choice, label="Select a VM Rental Plan", value="pay as you go", info="These options are from Azure's VM rental plans")
-                rental_plan_inp2.change(fn=calculate_tco, inputs=[model_inp2, rental_plan_inp2, out_diy2], outputs=out_diy2)
+                with gr.Column():
+                    with gr.Row():
+                        model_inp2 = gr.Dropdown(models, label="Select an AI Model", info="Open-source AI model used for your application")
+                
+                    with gr.Row() as vm2:
+                        with gr.Column():
+                            with gr.Row():
+                                vm_inp2 = gr.Dropdown(vm_choice, label="Select a Virtual Machine", info="Your options for this choice depend on the model you previously chose")
+                            with gr.Row(visible=False) as vm_info2:
+                                tokens_per_second2 = gr.Textbox(interactive=False, label="Token/s", info="To compute this value based on your model and VM choice, we chose an input length of 233 tokens.")
+                                vm_cost_per_hour2 = gr.Textbox(interactive=False, label="Cost/h ($) for the VM")
+                                                    
+                    with gr.Row() as use_case2:
+                        maxed_out2 = gr.Slider(minimum=0.01, value=80, label="% maxed out", info="percentage of how much your machine is maxed out")
+                        used2 = gr.Slider(minimum=0.01, value=50, label="% used", info="percentage of time your machine is used")
+                        rental_plan_inp2 = gr.Dropdown(vm_rental_choice, label="Select a VM Rental Plan", info="These options are from Azure's VM rental plans")
+                    
+                    model_inp2.change(fn=update_vm_choice, inputs=model_inp2, outputs=vm_inp2)
+                    vm_inp2.change(fn=token_per_s_and_cost, inputs=vm_inp2, outputs=[tokens_per_second_inp2, vm_cost_per_hour_inp2, tokens_per_second2, vm_cost_per_hour2])
+                    
+                    maxed_out2.change(fn=update_tco, inputs=[maxed_out2, used2, tokens_per_second_inp2, vm_cost_per_hour_inp2, rental_plan_inp2, out_diy2], outputs=out_diy2)
+                    used2.change(fn=update_tco, inputs=[maxed_out2, used2, tokens_per_second_inp2, vm_cost_per_hour_inp2, rental_plan_inp2, out_diy2], outputs=out_diy2)
+                    model_inp2.change(fn=update_tco, inputs=[maxed_out2, used2, tokens_per_second_inp2, vm_cost_per_hour_inp2, rental_plan_inp2, out_diy2], outputs=out_diy2)
+                    vm_inp2.change(fn=update_tco, inputs=[maxed_out2, used2, tokens_per_second_inp2, vm_cost_per_hour_inp2, rental_plan_inp2, out_diy2], outputs=out_diy2)
+                    rental_plan_inp2.change(fn=update_tco, inputs=[maxed_out2, used2, tokens_per_second_inp2, vm_cost_per_hour_inp2, rental_plan_inp2, out_diy2], outputs=out_diy2)
                       
             with gr.Row(visible=False) as text_saas_column2:            
                 gr.Markdown(description2)
@@ -220,8 +320,17 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
             
             with gr.Row(visible=False) as input_saas_column2:
                 model_provider_inp2 = gr.Dropdown(['OpenAI'], label="Model Provider", value="OpenAI", info="Choose an AI model provider you want to work with")
-                context_inp2 = gr.Dropdown(['4K context', '16K context'], label="Context", value="4K context", info="Number of tokens the model considers when processing text")
-                context_inp2.change(fn=calculate_tco_2, inputs=[model_provider_inp2, context_inp2, out_saas2], outputs=out_saas2)
+                context_inp2 = gr.Dropdown(['4K context', '16K context'], label="Context", info="Number of tokens the model considers when processing text")
+                
+                model_provider_inp2.change(fn=update_tco2, inputs=[model_provider_inp2, context_inp2, out_saas2], outputs=out_saas2)
+                context_inp2.change(fn=update_tco2, inputs=[model_provider_inp2, context_inp2, out_saas2], outputs=out_saas2)
+            
+            def show_vm_info():
+                            return {
+                                vm_info2: gr.update(visible=True),
+                            }
+
+            vm_inp2.change(show_vm_info, outputs=vm_info2)
             
             def submit2(solution_selection2):
                 if solution_selection2 == "Open-source":
@@ -248,7 +357,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
             solution_selection2.change(
                 submit2,
                 solution_selection2,
-                [out_diy2, out_saas2, formula_diy2, formula_saas2, title_column2, text_diy_column2, text_saas_column2, model_inp2, rental_plan_inp2, model_provider_inp2, context_inp2, input_diy_column2, input_saas_column2],
+                [vm2, vm_info2, vm_inp2, maxed_out2, used2, out_diy2, out_saas2, formula_diy2, formula_saas2, title_column2, text_diy_column2, text_saas_column2, model_inp2, rental_plan_inp2, model_provider_inp2, context_inp2, input_diy_column2, input_saas_column2],
             )
     
     with gr.Row():
@@ -261,11 +370,17 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
                 plot = gr.BarPlot(vertical=False, title="Comparison", y_title="Cost/token ($)", width=500, interactive=True)
                 
             context_inp.change(fn=update_plot, inputs=[out_diy2, out_saas], outputs=plot)
+            maxed_out2.change(fn=update_plot, inputs=[out_diy2, out_saas], outputs=plot)
+            used2.change(fn=update_plot, inputs=[out_diy2, out_saas], outputs=plot)
+            vm_inp2.change(fn=update_plot, inputs=[out_diy2, out_saas], outputs=plot)
             model_provider_inp.change(fn=update_plot, inputs=[out_diy2, out_saas], outputs=plot)
             rental_plan_inp2.change(fn=update_plot, inputs=[out_diy2, out_saas], outputs=plot)
             model_inp2.change(fn=update_plot, inputs=[out_diy2, out_saas], outputs=plot)
             
             context_inp2.change(fn=update_plot, inputs=[out_diy, out_saas2], outputs=plot)
+            vm_inp.change(fn=update_plot, inputs=[out_diy, out_saas2], outputs=plot)
+            maxed_out.change(fn=update_plot, inputs=[out_diy, out_saas2], outputs=plot)
+            used.change(fn=update_plot, inputs=[out_diy, out_saas2], outputs=plot)
             model_provider_inp2.change(fn=update_plot, inputs=[out_diy, out_saas2], outputs=plot)
             rental_plan_inp.change(fn=update_plot, inputs=[out_diy, out_saas2], outputs=plot)
             model_inp.change(fn=update_plot, inputs=[out_diy, out_saas2], outputs=plot)