Update

app-ngrok.py

@@ -126,18 +126,18 @@ with gr.Blocks(theme='gradio/soft') as demo:
     note = gr.HTML("""<p style="font-size: 12px; text-align: center">⚠️ Should take 30-60s to generate</p>""")
     input_text = gr.Textbox(lines=3, placeholder='Write your question here...', label='NL Input')
     db_info = gr.Textbox(lines=4, placeholder='Example: | table_01 : column_01 , column_02 | table_02 : column_01 , column_02 | ...', label='Database Info')
+    format_sql = gr.Checkbox(label="Format SQL + Remove Skeleton", value=True, interactive=True)
+    
+    # Generate button UI element
+    run_button = gr.Button("Generate SQL", variant="primary")
 
     with gr.Accordion("Options", open=False):
         temperature = gr.Slider(label="Temperature", minimum=0.0, maximum=1.0, value=0.2, step=0.1)
         top_p = gr.Slider(label="Top-p (nucleus sampling)", minimum=0.0, maximum=1.0, value=0.9, step=0.01)
         top_k = gr.Slider(label="Top-k", minimum=0, maximum=200, value=0, step=1)
         repetition_penalty = gr.Slider(label="Repetition Penalty", minimum=1.0, maximum=2.0, value=1.08, step=0.01)
-        format_sql = gr.Checkbox(label="Format SQL + Remove Skeleton", value=True, interactive=True)
         stop_sequence = gr.Textbox(lines=1, value="Explanation,Note", label='Extra Stop Sequence')
     
-    # Generate button UI element
-    run_button = gr.Button("Generate SQL", variant="primary")
-    
     ## Add statement saying that inputs/outpus are sent to firebase
     info = gr.HTML(f"""
         <p>🌐 Leveraging the <a href='https://huggingface.co/{quantized_model}'><strong>4-bit GGML version</strong></a> of <a href='https://huggingface.co/{merged_model}'><strong>{merged_model}</strong></a> model.</p>

test.py

@@ -1,52 +1,74 @@
 import gradio as gr
 
-def bot(input_message: str, db_info="", temperature=0.1, top_p=0.9, top_k=0, repetition_penalty=1.08):
+def bot(input_message: str, db_info="", temperature=0.1, top_p=0.9, top_k=0, repetition_penalty=1.08, format_sql=True, stop_sequence="Explanation,Note", log=True):
     # For the stripped down version, let's just return a preset output
     final_query = "| stadium : stadium_id , location , name , capacity , highest , lowest , average | singer : singer_id , name , country , song_name , song_release_year , age , is_male | concert : concert_id , concert_name , theme , stadium_id , year | singer_in_concert : concert_id , singer_id | concert.stadium_id = stadium.stadium_id | singer_in_concert.singer_id = singer.singer_id | singer_in_concert.concert_id = concert.concert_id |" 
     final_query_markdown = f"{final_query}"
     return final_query_markdown
 
+# Gradio UI Code
 with gr.Blocks(theme='gradio/soft') as demo:
+    # Elements stack vertically by default just define elements in order you want them to stack
     header = gr.HTML("""
         <h1 style="text-align: center">SQL Skeleton WizardCoder Demo</h1>
-        <h3 style="text-align: center">🧙‍♂️ Generate SQL queries from Natural Language 🧙‍♂️</h3>
+        <h3 style="text-align: center">🕷️☠️🧙‍♂️ Generate SQL queries from Natural Language 🕷️☠️🧙‍♂️</h3>
     """)
 
     output_box = gr.Code(label="Generated SQL", lines=2, interactive=True)
+    note = gr.HTML("""<p style="font-size: 12px; text-align: center">⚠️ Should take 30-60s to generate</p>""")
     input_text = gr.Textbox(lines=3, placeholder='Write your question here...', label='NL Input')
     db_info = gr.Textbox(lines=4, placeholder='Example: | table_01 : column_01 , column_02 | table_02 : column_01 , column_02 | ...', label='Database Info')
+    format_sql = gr.Checkbox(label="Format SQL + Remove Skeleton", value=True, interactive=True)
+    
+    # Generate button UI element
+    run_button = gr.Button("Generate SQL", variant="primary")
 
-    with gr.Accordion("Hyperparameters", open=False):
-        temperature = gr.Slider(label="Temperature", minimum=0.0, maximum=1.0, value=0.5, step=0.1)
+    with gr.Accordion("Options", open=False):
+        temperature = gr.Slider(label="Temperature", minimum=0.0, maximum=1.0, value=0.2, step=0.1)
         top_p = gr.Slider(label="Top-p (nucleus sampling)", minimum=0.0, maximum=1.0, value=0.9, step=0.01)
         top_k = gr.Slider(label="Top-k", minimum=0, maximum=200, value=0, step=1)
         repetition_penalty = gr.Slider(label="Repetition Penalty", minimum=1.0, maximum=2.0, value=1.08, step=0.01)
-        
-    run_button = gr.Button("Generate SQL", variant="primary")
+        stop_sequence = gr.Textbox(lines=1, value="Explanation,Note", label='Extra Stop Sequence')
     
-    with gr.Accordion("Examples", open=True):
+    ## Add statement saying that inputs/outpus are sent to firebase
+    info = gr.HTML(f"""
+        <p>🌐 Leveraging the <a href='https://huggingface.co/{quantized_model}'><strong>4-bit GGML version</strong></a> of <a href='https://huggingface.co/{merged_model}'><strong>{merged_model}</strong></a> model.</p>
+        <p>🔗 How it's made: <a href='https://huggingface.co/{initial_model}'><strong>{initial_model}</strong></a> was finetuned to create <a href='https://huggingface.co/{lora_model}'><strong>{lora_model}</strong></a>, then merged together to create <a href='https://huggingface.co/{merged_model}'><strong>{merged_model}</strong></a>.</p>
+        <p>📉 Fine-tuning was performed using QLoRA techniques on the <a href='https://huggingface.co/datasets/{dataset}'><strong>{dataset}</strong></a> dataset. You can view training metrics on the <a href='https://huggingface.co/{lora_model}'><strong>QLoRa adapter HF Repo</strong></a>.</p>
+        <p>📊 All inputs/outputs are logged to Firebase to see how the model is doing.</a></p>
+    """)
+
+    examples = gr.Examples([
+        ["What is the average, minimum, and maximum age of all singers from France?", "| stadium : stadium_id , location , name , capacity , highest , lowest , average | singer : singer_id , name , country , song_name , song_release_year , age , is_male | concert : concert_id , concert_name , theme , stadium_id , year | singer_in_concert : concert_id , singer_id | concert.stadium_id = stadium.stadium_id | singer_in_concert.singer_id = singer.singer_id | singer_in_concert.concert_id = concert.concert_id |"],
+        ["How many students have dogs?", "| student : stuid , lname , fname , age , sex , major , advisor , city_code | has_pet : stuid , petid | pets : petid , pettype , pet_age , weight | has_pet.stuid = student.stuid | has_pet.petid = pets.petid | pets.pettype = 'Dog' |"],
+    ], inputs=[input_text, db_info, temperature, top_p, top_k, repetition_penalty, format_sql, stop_sequence], fn=generate, cache_examples=False if platform.system() == "Windows" or platform.system() == "Darwin" else True, outputs=output_box)
+
+    with gr.Accordion("More Examples", open=False):
         examples = gr.Examples([
-            ["What is the average, minimum, and maximum age for all French singers?", "| stadium : stadium_id , location , name , capacity , highest , lowest , average | singer : singer_id , name , country , song_name , song_release_year , age , is_male | concert : concert_id , concert_name , theme , stadium_id , year | singer_in_concert : concert_id , singer_id | concert.stadium_id = stadium.stadium_id | singer_in_concert.singer_id = singer.singer_id | singer_in_concert.concert_id = concert.concert_id |"],
+            ["What is the average weight of pets of all students?", "| student : stuid , lname , fname , age , sex , major , advisor , city_code | has_pet : stuid , petid | pets : petid , pettype , pet_age , weight | has_pet.stuid = student.stuid | has_pet.petid = pets.petid |"],
+            ["How many male singers performed in concerts in the year 2023?", "| stadium : stadium_id , location , name , capacity , highest , lowest , average | singer : singer_id , name , country , song_name , song_release_year , age , is_male | concert : concert_id , concert_name , theme , stadium_id , year | singer_in_concert : concert_id , singer_id | concert.stadium_id = stadium.stadium_id | singer_in_concert.singer_id = singer.singer_id | singer_in_concert.concert_id = concert.concert_id |"],
+            ["For students who have pets, how many pets does each student have? List their ids instead of names.", "| student : stuid , lname , fname , age , sex , major , advisor , city_code | has_pet : stuid , petid | pets : petid , pettype , pet_age , weight | has_pet.stuid = student.stuid | has_pet.petid = pets.petid |"],
             ["Show location and name for all stadiums with a capacity between 5000 and 10000.", "| stadium : stadium_id , location , name , capacity , highest , lowest , average | singer : singer_id , name , country , song_name , song_release_year , age , is_male | concert : concert_id , concert_name , theme , stadium_id , year | singer_in_concert : concert_id , singer_id | concert.stadium_id = stadium.stadium_id | singer_in_concert.singer_id = singer.singer_id | singer_in_concert.concert_id = concert.concert_id |"],
             ["What are the number of concerts that occurred in the stadium with the largest capacity ?", "| stadium : stadium_id , location , name , capacity , highest , lowest , average | singer : singer_id , name , country , song_name , song_release_year , age , is_male | concert : concert_id , concert_name , theme , stadium_id , year | singer_in_concert : concert_id , singer_id | concert.stadium_id = stadium.stadium_id | singer_in_concert.singer_id = singer.singer_id | singer_in_concert.concert_id = concert.concert_id |"],
-            ["How many male singers performed in concerts in the year 2023?", "| stadium : stadium_id , location , name , capacity , highest , lowest , average | singer : singer_id , name , country , song_name , song_release_year , age , is_male | concert : concert_id , concert_name , theme , stadium_id , year | singer_in_concert : concert_id , singer_id | concert.stadium_id = stadium.stadium_id | singer_in_concert.singer_id = singer.singer_id | singer_in_concert.concert_id = concert.concert_id |"],
-            ["List the names of all singers who performed in a concert with the theme 'Rock'", "| stadium : stadium_id , location , name , capacity , highest , lowest , average | singer : singer_id , name , country , song_name , song_release_year , age , is_male | concert : concert_id , concert_name , theme , stadium_id , year | singer_in_concert : concert_id , singer_id | concert.stadium_id = stadium.stadium_id | singer_in_concert.singer_id = singer.singer_id | singer_in_concert.concert_id = concert.concert_id |"]
-        ], inputs=[input_text, db_info, temperature, top_p, top_k, repetition_penalty], fn=bot)
-
-    bitsandbytes_model = "richardr1126/spider-skeleton-wizard-coder-8bit"
-    merged_model = "richardr1126/spider-skeleton-wizard-coder-merged"
-    initial_model = "WizardLM/WizardCoder-15B-V1.0"
-    finetuned_model = "richardr1126/spider-skeleton-wizard-coder-qlora"
-    dataset = "richardr1126/spider-skeleton-context-instruct"
-    
-    footer = gr.HTML(f"""
-        <p>🛠️ If you want you can <strong>duplicate this Space</strong>, then change the HF_MODEL_REPO spaces env varaible to use any Transformers model.</p>
-        <p>🌐 Leveraging the <a href='https://huggingface.co/{bitsandbytes_model}'><strong>bitsandbytes 8-bit version</strong></a> of <a href='https://huggingface.co/{merged_model}'><strong>{merged_model}</strong></a> model.</p>
-        <p>🔗 How it's made: <a href='https://huggingface.co/{initial_model}'><strong>{initial_model}</strong></a> was finetuned to create <a href='https://huggingface.co/{finetuned_model}'><strong>{finetuned_model}</strong></a>, then merged together to create <a href='https://huggingface.co/{merged_model}'><strong>{merged_model}</strong></a>.</p>
-        <p>📉 Fine-tuning was performed using QLoRA techniques on the <a href='https://huggingface.co/datasets/{dataset}'><strong>{dataset}</strong></a> dataset. You can view training metrics on the <a href='https://huggingface.co/{finetuned_model}'><strong>QLoRa adapter HF Repo</strong></a>.</p>
-    """)
+            ["Which student has the oldest pet?", "| student : stuid , lname , fname , age , sex , major , advisor , city_code | has_pet : stuid , petid | pets : petid , pettype , pet_age , weight | has_pet.stuid = student.stuid | has_pet.petid = pets.petid |"],
+            ["List the names of all singers who performed in a concert with the theme 'Rock'", "| stadium : stadium_id , location , name , capacity , highest , lowest , average | singer : singer_id , name , country , song_name , song_release_year , age , is_male | concert : concert_id , concert_name , theme , stadium_id , year | singer_in_concert : concert_id , singer_id | concert.stadium_id = stadium.stadium_id | singer_in_concert.singer_id = singer.singer_id | singer_in_concert.concert_id = concert.concert_id |"],
+            ["List all students who don't have pets.", "| student : stuid , lname , fname , age , sex , major , advisor , city_code | has_pet : stuid , petid | pets : petid , pettype , pet_age , weight | has_pet.stuid = student.stuid | has_pet.petid = pets.petid |"],
+        ], inputs=[input_text, db_info, temperature, top_p, top_k, repetition_penalty, format_sql, stop_sequence], fn=bot, cache_examples=False, outputs=output_box)
+
 
+    readme_content = requests.get(f"https://huggingface.co/{merged_model}/raw/main/README.md").text
+    readme_content = re.sub('---.*?---', '', readme_content, flags=re.DOTALL) #Remove YAML front matter
 
-    run_button.click(fn=bot, inputs=[input_text, db_info, temperature, top_p, top_k, repetition_penalty], outputs=output_box, api_name="txt2sql")
+    with gr.Accordion("📖 Model Readme", open=True):
+        readme = gr.Markdown(
+            readme_content,
+        )
+    
+    with gr.Accordion("More Options:", open=False):
+        log = gr.Checkbox(label="Log to Firebase", value=True, interactive=True)
+    
+    # When the button is clicked, call the generate function, inputs are taken from the UI elements, outputs are sent to outputs elements
+    run_button.click(fn=bot, inputs=[input_text, db_info, temperature, top_p, top_k, repetition_penalty, format_sql, stop_sequence, log], outputs=output_box, api_name="txt2sql")
+    
 
-demo.queue(concurrency_count=1, max_size=10).launch()
+demo.queue(concurrency_count=1, max_size=20).launch(debug=True)