Use my hosted API

README.md

@@ -5,7 +5,7 @@ colorFrom: gray
 colorTo: purple
 sdk: gradio
 sdk_version: 3.37.0
-app_file: app.py
+app_file: app-kobold.py
 pinned: true
 duplicated_from: richardr1126/natsql-wizardcoder-demo
 license: bigcode-openrail-m

app-kobold.py

@@ -0,0 +1,105 @@
+import os
+import gradio as gr
+import sqlparse
+import requests
+from time import sleep
+
+def format(text):
+    # Split the text by "|", and get the last element in the list which should be the final query
+    try:
+        final_query = text.split("|")[1].strip()
+    except Exception:
+        final_query = text
+
+    try:
+        # Attempt to format SQL query using sqlparse
+        formatted_query = sqlparse.format(final_query, reindent=True, keyword_case='upper')
+    except Exception:
+        # If formatting fails, use the original, unformatted query
+        formatted_query = final_query
+
+    # Convert SQL to markdown (not required, but just to show how to use the markdown module)
+    final_query_markdown = f"{formatted_query}"
+
+    return final_query_markdown
+
+
+def bot(input_message: str, db_info="", temperature=0.1, top_p=0.9, top_k=0, repetition_penalty=1.08):
+    # Format the user's input message
+    messages = f"Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.\n\n### Instruction:\n\nConvert text to sql: {input_message} {db_info}\n\n### Response:\n\n"
+
+    url = "https://https://e9f4be879d38-8269039109365193683.ngrok-free.app/api/v1/generate"
+    payload = {
+        "prompt": messages,
+        "temperature": temperature,
+        "top_p": top_p,
+        "top_k": top_k,
+        "top_a": 0,
+        "n": 1,
+        "max_context_length": 2048,
+        "max_length": 512,
+        "rep_pen": repetition_penalty,
+        "sampler_order": [6,0,1,3,4,2,5],
+        "stop_sequence": ["###", "Result"],
+    }
+    headers = {"Content-Type": "application/json"}
+
+    for _ in range(3):
+        try:
+            response = requests.post(url, json=payload, headers=headers)
+            response_text = response.json()["results"][0]["text"]
+            response_text = response_text.replace("\n", "").replace("\t", " ")
+            if response_text and response_text[-1] == ".":
+                response_text = response_text[:-1]
+
+            return format(response_text)
+            
+        except Exception as e:
+            print(f'Error occurred: {str(e)}')
+            print('Waiting for 10 seconds before retrying...')
+            sleep(10)
+
+with gr.Blocks(theme='gradio/soft') as demo:
+    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>
+    """)
+
+    output_box = gr.Code(label="Generated SQL", lines=2, interactive=True)
+    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')
+
+    with gr.Accordion("Hyperparameters", open=False):
+        temperature = gr.Slider(label="Temperature", minimum=0.0, maximum=1.0, value=0.5, 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")
+    
+    with gr.Accordion("Examples", open=True):
+        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 |"],
+            ["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)
+
+    quantized_model = "richardr1126/spider-skeleton-wizard-coder-ggml"
+    merged_model = "richardr1126/spider-skeleton-wizard-coder-merged"
+    initial_model = "WizardLM/WizardCoder-15B-V1.0"
+    lora_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 GGML model.</p>
+        <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>
+    """)
+
+
+    run_button.click(fn=bot, inputs=[input_text, db_info, temperature, top_p, top_k, repetition_penalty], outputs=output_box, api_name="txt2sql")
+
+demo.queue(concurrency_count=1, max_size=10).launch(debug=True)