Spaces:
Running
Running
| """LaunchReady Space: will this model fit on this GPU? A thin Gradio page over logic.py. | |
| Stays anonymous even if a token is added to the Space later: a public page holding a token would | |
| let anyone read gated repos through it. Both lines must run before huggingface_hub is imported. | |
| """ | |
| import os | |
| os.environ.pop("HF_TOKEN", None) | |
| os.environ["HF_HUB_DISABLE_IMPLICIT_TOKEN"] = "1" | |
| os.environ["GRADIO_ANALYTICS_ENABLED"] = "False" | |
| import gradio as gr # noqa: E402 | |
| from logic import ( # noqa: E402 | |
| DEFAULT_CONTEXT, | |
| DEFAULT_USERS, | |
| EXAMPLES, | |
| LIGHTNING, | |
| check, | |
| how_calculated_text, | |
| load_gpus, | |
| measured_text, | |
| ) | |
| INTRO = """# LaunchReady | |
| **Will this open model run on my NVIDIA GPU?** | |
| Think of the model as a stack of papers and your GPU's memory as a desk. LaunchReady works out | |
| whether the stack fits on the desk, with room left for everyone using it, before you download it. | |
| Enter a model from Hugging Face and how you'd use it. Nothing is downloaded, and every result is a | |
| projection, not a measurement. | |
| """ | |
| # Gradio breaks table cells mid-word on a phone; keep words whole and scroll the table instead. | |
| CSS = """ | |
| .md table { display: block; overflow-x: auto; } | |
| .md th, .md td { word-break: normal; overflow-wrap: normal; } | |
| """ | |
| def run(repo_id, gpu_id, context, users): | |
| page = check(repo_id, gpu_id, context, users) | |
| return page.verdict, page.memory, page.warnings, page.variants | |
| gpus = load_gpus() | |
| with gr.Blocks(title="LaunchReady") as demo: | |
| gr.Markdown(INTRO) | |
| repo = gr.Textbox( | |
| label="Model", | |
| value=LIGHTNING, | |
| info="A Hugging Face model id or link, like Qwen/Qwen3-4B", | |
| lines=1, | |
| max_lines=1, | |
| ) | |
| with gr.Row(): | |
| gpu = gr.Dropdown(label="GPU", choices=[(g.name, g.id) for g in gpus], value="rtx-5090") | |
| context = gr.Number( | |
| label="Context length", | |
| value=DEFAULT_CONTEXT, | |
| precision=0, | |
| minimum=1, | |
| info="Tokens per conversation", | |
| ) | |
| users = gr.Number( | |
| label="Number of users", | |
| value=DEFAULT_USERS, | |
| precision=0, | |
| minimum=1, | |
| info="At the same time", | |
| ) | |
| inputs = [repo, gpu, context, users] | |
| example_row = gr.Column() | |
| button = gr.Button("Check", variant="primary") | |
| verdict = gr.Markdown("Press **Check**, or pick an example above.") | |
| memory = gr.Markdown() | |
| warnings = gr.Markdown() | |
| variants = gr.Markdown() | |
| outputs = [verdict, memory, warnings, variants] | |
| button.click(run, inputs, outputs) | |
| repo.submit(run, inputs, outputs) | |
| with example_row: | |
| gr.Examples( | |
| examples=EXAMPLES, | |
| inputs=inputs, | |
| outputs=outputs, | |
| fn=run, | |
| run_on_click=True, | |
| cache_examples=False, | |
| label="Examples: click one to run it", | |
| ) | |
| gr.Markdown(how_calculated_text(gpus)) | |
| gr.Markdown(measured_text()) | |
| if __name__ == "__main__": | |
| demo.launch(css=CSS) | |