Merge branch 'main' of https://huggingface.co/spaces/Intel/powered_by_intel_llm_leaderboard

app.py

@@ -3,12 +3,6 @@ import requests
 import os
 
 import gradio
-
-# work around due to HF Spaces bug
-#if gradio.__version__ != '4.16.0':
-#    os.system("pip uninstall -y gradio")
-#    os.system("pip install gradio==4.16.0")
-
 import gradio as gr
 
 from info.train_a_model import (
@@ -77,62 +71,62 @@ with demo:
         except Exception as e:
             return f"❌Failed to submit due to an error: {str(e)}"
     
-    with gr.Accordion("Chat with Top Models on the Leaderboard Here 💬", open=False):
-        
-        chat_model_dropdown = gr.Dropdown(
-                        choices=VALIDATED_CHAT_MODELS,
-                        label="Select a leaderboard model to chat with. ",
-                        multiselect=False,
-                        value=VALIDATED_CHAT_MODELS[0],
-                        interactive=True,
-                    )
-        
-        #chat_model_selection = chat_model_dropdown.value
-        chat_model_selection = 'yuriachermann/My_AGI_llama_2_7B'
-        
-        def call_api_and_stream_response(query, chat_model):
-            """
-            Call the API endpoint and yield characters as they are received.
-            This function simulates streaming by yielding characters one by one.
-            """
-            url = inference_endpoint_url
-            params = {"query": query, "selected_model": chat_model}
-            with requests.get(url, json=params, stream=True) as r:  # Use params for query parameters
-                for chunk in r.iter_content(chunk_size=1):
-                    if chunk:
-                        yield chunk.decode()
-
-        def get_response(query, history):
-            """
-            Wrapper function to call the streaming API and compile the response.
-            """
-            response = ''
-            for char in call_api_and_stream_response(query, chat_model=chat_model_selection):
-                if char == '<':  # This is stopping condition; adjust as needed.
-                    break
-                response += char
-                yield [(f"🤖 Response from LLM: {chat_model_selection}", response)]  # Correct format for Gradio Chatbot
+    #with gr.Accordion("Chat with Top Models on the Leaderboard Here 💬", open=False):
+    #    
+    #    chat_model_dropdown = gr.Dropdown(
+    #                    choices=VALIDATED_CHAT_MODELS,
+    #                    label="Select a leaderboard model to chat with. ",
+    #                    multiselect=False,
+    #                    value=VALIDATED_CHAT_MODELS[0],
+    #                    interactive=True,
+    #                )
+    #    
+    #    #chat_model_selection = chat_model_dropdown.value
+    #    chat_model_selection = 'yuriachermann/My_AGI_llama_2_7B'
+    #    
+    #    def call_api_and_stream_response(query, chat_model):
+    #        """
+    #        Call the API endpoint and yield characters as they are received.
+    #        This function simulates streaming by yielding characters one by one.
+    #        """
+    #        url = inference_endpoint_url
+    #        params = {"query": query, "selected_model": chat_model}
+    #        with requests.get(url, json=params, stream=True) as r:  # Use params for query parameters
+    #            for chunk in r.iter_content(chunk_size=1):
+    #                if chunk:
+    #                    yield chunk.decode()
 #
-
-        chatbot = gr.Chatbot()
-        msg = gr.Textbox()
-        submit = gr.Button("Submit")
-        clear = gr.Button("Clear")
-        def user(user_message, history):
-            return "", history + [[user_message, None]]
-        def clear_chat(*args):
-            return []  # Returning an empty list to signify clearing the chat, adjust as per Gradio's capabilities
-        submit.click(
-            fn=get_response,
-            inputs=[msg, chatbot],
-            outputs=chatbot
-        )
-        clear.click(
-            fn=clear_chat,
-            inputs=None,
-            outputs=chatbot
-        )
-        
+    #    def get_response(query, history):
+    #        """
+    #        Wrapper function to call the streaming API and compile the response.
+    #        """
+    #        response = ''
+    #        for char in call_api_and_stream_response(query, chat_model=chat_model_selection):
+    #            if char == '<':  # This is stopping condition; adjust as needed.
+    #                break
+    #            response += char
+    #            yield [(f"🤖 Response from LLM: {chat_model_selection}", response)]  # Correct format for Gradio Chatbot
+##
+#
+    #    chatbot = gr.Chatbot()
+    #    msg = gr.Textbox()
+    #    submit = gr.Button("Submit")
+    #    clear = gr.Button("Clear")
+    #    def user(user_message, history):
+    #        return "", history + [[user_message, None]]
+    #    def clear_chat(*args):
+    #        return []  # Returning an empty list to signify clearing the chat, adjust as per Gradio's capabilities
+    #    submit.click(
+    #        fn=get_response,
+    #        inputs=[msg, chatbot],
+    #        outputs=chatbot
+    #    )
+    #    clear.click(
+    #        fn=clear_chat,
+    #        inputs=None,
+    #        outputs=chatbot
+    #    )
+    #    
 
     with gr.Tabs(elem_classes="tab-buttons") as tabs:
         with gr.TabItem("🏆 LLM Leaderboard", elem_id="llm-benchmark-table", id=0):
@@ -152,10 +146,10 @@ with demo:
                                      value=["No Affiliation","Intel Innovator","Student Ambassador","Intel Liftoff", "Intel Engineering", "Other"])
                     
                 with gr.Column():
-                    filter_size = gr.CheckboxGroup(choices=[1,2,3,5,7,13,35,60,70,100],
+                    filter_size = gr.CheckboxGroup(choices=[1,2,3,5,7,8,13,35,60,70,100],
                                      label="Model Sizes (Billion of Parameters)",
                                      elem_id="parameter_size",
-                                     value=[1,2,3,5,7,13,35,60,70,100])
+                                     value=[1,2,3,5,7,8,13,35,60,70,100])
                     filter_precision = gr.CheckboxGroup(choices=["fp32","fp16","bf16","int8","fp8", "int4"],
                                      label="Model Precision",
                                      elem_id="precision",
@@ -187,7 +181,7 @@ with demo:
             initial_filtered_df = update_df(["Gaudi","Xeon","GPU Max","Arc GPU","Core Ultra"], 
                                 ["Intel Developer Cloud","AWS","Azure","Google Cloud Platform","Local"], 
                                 ["No Affiliation","Intel Innovator","Student Ambassador","Intel Liftoff", "Intel Engineering", "Other"], 
-                                [1,2,3,5,7,13,35,60,70,100], 
+                                [1,2,3,5,7,8,13,35,60,70,100], 
                                 ["fp32","fp16","bf16","int8","fp8", "int4"], 
                                 ["pretrained","fine-tuned","chat-models","merges/moerges"])