Update models.py

models.py

@@ -112,11 +112,21 @@ class OpenSourceLlama2Model(BaseTCOModel):
         
         def on_model_change(model):
             if model == "Llama 2 7B":
-                return [gr.Dropdown.update(choices=vm_choices), gr.Markdown.update(visible=True), gr.Markdown.update(visible=False)]
+                return [gr.Dropdown.update(choices=vm_choices), 
+                        gr.Markdown.update(value="To see the script used to benchmark the Llama2-7B model, [click here](https://example.com/script)"), 
+                        gr.Number.update(value=3.6730),
+                        gr.Number.update(value=694.38),
+                        gr.Number.update(visible=True)
+                        ]
             else:
                 not_supported_vm = ["1x Nvidia A100 (Azure NC24ads A100 v4)", "2x Nvidia A100 (Azure NC48ads A100 v4)"]
                 choices = [x for x in vm_choices if x not in not_supported_vm]
-                return [gr.Dropdown.update(choices=choices), gr.Markdown.update(visible=False), gr.Markdown.update(visible=True)]
+                return [gr.Dropdown.update(choices=choices, value="4x Nvidia A100 (Azure NC48ads A100 v4)"), 
+                        gr.Markdown.update(value="To see the benchmark results used for the Llama2-70B model, [click here](https://www.cursor.so/blog/llama-inference#user-content-fn-llama-paper)"),
+                        gr.Number.update(value=14.692),
+                        gr.Number.update(value=18.6),
+                        gr.Number.update(visible=False)
+                        ]
 
         def on_vm_change(model, vm):
             # TO DO: load info from CSV
@@ -144,10 +154,9 @@ class OpenSourceLlama2Model(BaseTCOModel):
                                            )
         self.input_length = gr.Number(233, label="Average number of input tokens", info="This is the number of input tokens used when the model was benchmarked to get the number of tokens/second it processes",
                                       interactive=False, visible=False)
-        self.info_7B = gr.Markdown("To see the script used to benchmark the Llama2-7B model, [click here](https://example.com/script)", interactive=False, visible=False)
-        self.info_70B = gr.Markdown("To see the benchmark results used for the Llama2-70B model, [click here](https://www.cursor.so/blog/llama-inference#user-content-fn-llama-paper)", interactive=False, visible=False)
+        self.info = gr.Markdown("To see the script used to benchmark the Llama2-7B model, [click here](https://example.com/script)", interactive=False, visible=False)
         
-        self.model.change(on_model_change, inputs=self.model, outputs=[self.vm, self.info_7B, self.info_70B])
+        self.model.change(on_model_change, inputs=self.model, outputs=[self.vm, self.info, self.vm_cost_per_hour, self.tokens_per_second, self.input_length])
         self.vm.change(on_vm_change, inputs=[self.model, self.vm], outputs=[self.vm_cost_per_hour, self.tokens_per_second])
         self.maxed_out = gr.Slider(minimum=0.01, value=50., step=0.01, label="% maxed out", 
                                    info="How much the GPU is fully used",