|
import gradio as gr |
|
from evaluation_logic import run_evaluation, AVAILABLE_PROMPT_FORMATS |
|
|
|
def gradio_run_evaluation(inference_api, model_name, prompt_format): |
|
output = [] |
|
for result in run_evaluation(inference_api, str(model_name).strip(), prompt_format): |
|
output.append(result) |
|
yield "\n".join(output) |
|
|
|
with gr.Blocks(gr.themes.Soft()) as demo: |
|
gr.Markdown("# DuckDB SQL Evaluation App") |
|
|
|
inference_api = gr.Dropdown( |
|
label="Inference API", |
|
choices=['openrouter'], |
|
value="openrouter" |
|
) |
|
model_name = gr.Textbox(label="Model Name (e.g., qwen/qwen-2.5-72b-instruct)") |
|
gr.Markdown("[View OpenRouter Models](https://openrouter.ai/models?order=top-weekly)") |
|
|
|
prompt_format = gr.Dropdown( |
|
label="Prompt Format", |
|
choices=['duckdbinst', 'duckdbinstgraniteshort'], |
|
value="duckdbinstgraniteshort" |
|
) |
|
gr.Examples( |
|
examples=[ |
|
["openrouter", "qwen/qwen-2.5-72b-instruct", "duckdbinst"], |
|
["openrouter", "meta-llama/llama-3.2-3b-instruct:free", "duckdbinstgraniteshort"], |
|
["openrouter", "mistralai/mistral-nemo", "duckdbinst"], |
|
], |
|
inputs=[inference_api, model_name, prompt_format], |
|
) |
|
|
|
start_btn = gr.Button("Start Evaluation") |
|
output = gr.Textbox(label="Output", lines=20) |
|
|
|
start_btn.click(fn=gradio_run_evaluation, inputs=[inference_api, model_name, prompt_format], outputs=output) |
|
|
|
demo.queue().launch() |
