app.py

import os, glob
import json
from datetime import datetime, timezone
from dataclasses import dataclass
from datasets import load_dataset, Dataset
import pandas as pd
import gradio as gr
from huggingface_hub import HfApi, snapshot_download, ModelInfo, list_models
from enum import Enum


OWNER = "EnergyStarAI"
COMPUTE_SPACE = f"{OWNER}/launch-computation-example"


TOKEN = os.environ.get("DEBUG")
API = HfApi(token=TOKEN)



task_mappings = {'automatic speech recognition':'automatic-speech-recognition', 'Object Detection': 'object-detection', 'Text Classification': 'text-classification', 
                 'Image to Text':'image-to-text', 'Question Answering':'question-answering', 'Text Generation': 'text-generation',
                 'Image Classification':'image-classification', 'Sentence Similarity': 'sentence-similarity',
                 'Image Generation':'image-generation', 'Summarization':'summarization'}

@dataclass
class ModelDetails:
    name: str
    display_name: str = ""
    symbol: str = "" # emoji

def start_compute_space():
    API.restart_space(COMPUTE_SPACE)  
    return f"Okay! {COMPUTE_SPACE} should be running now!"


def get_model_size(model_info: ModelInfo):
    """Gets the model size from the configuration, or the model name if the configuration does not contain the information."""
    try:
        model_size = round(model_info.safetensors["total"] / 1e9, 3)
    except (AttributeError, TypeError):
        return 0  # Unknown model sizes are indicated as 0, see NUMERIC_INTERVALS in app.py
    return model_size


def add_new_eval(
    repo_id: str,
    task: str,
):
    model_owner = repo_id.split("/")[0]
    model_name = repo_id.split("/")[1]  
    model_list=[]
    current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
    requests= load_dataset("EnergyStarAI/requests_debug", split="test", token=TOKEN)
    requests_dset = requests.to_pandas()
    model_list= requests_dset[requests_dset['status'] == 'COMPLETED']['model'].tolist()
    task_models = list(API.list_models(filter=task_mappings[task]))
    task_model_names = [m.id for m in task_models]
    if repo_id in model_list:
        return 'This model has already been run!'
    elif repo_id not in task_model_names:
        return "This model isn't compatible with the chosen task! Pick a different model-task combination"
    else:
        # Is the model info correctly filled?
        try:
            model_info = API.model_info(repo_id=repo_id)
        except Exception:
            return "Could not find information for model %s" % (model)
            
        model_size = get_model_size(model_info=model_info)
        
        print("Adding request")
    
        
        request_dict = {
            "model": repo_id,
            "status": "PENDING",
            "submitted_time": pd.to_datetime(current_time),
            "task": task,
            "likes": model_info.likes,
            "params": model_size,
            "leaderboard_version": "v0",}
            #"license": license,
            #"private": False,
        #}
    
        print("Writing out request file to dataset")
        df_request_dict = pd.DataFrame([request_dict])
        print(df_request_dict)
        df_final = pd.concat([requests_dset, df_request_dict], ignore_index=True)
        updated_dset =Dataset.from_pandas(df_final)
        updated_dset.push_to_hub("EnergyStarAI/requests_debug", split="test", token=TOKEN)
        
        print("Starting compute space at %s " % COMPUTE_SPACE)
        return start_compute_space()

def print_existing_models():
    requests= load_dataset("EnergyStarAI/requests_debug", split="test", token=TOKEN)
    requests_dset = requests.to_pandas()
    model_list= requests_dset[requests_dset['status'] == 'COMPLETED']
    return model_list[['model','task']]   

def get_leaderboard_models():
    path = r'leaderboard_v0_data/energy'
    filenames = glob.glob(path + "/*.csv")
    data = []
    for filename in filenames:
        data.append(pd.read_csv(filename))
    leaderboard_data = pd.concat(data, ignore_index=True)
    return leaderboard_data[['model','task']]


with gr.Blocks() as demo:
    gr.Markdown("# Energy Star Submission Portal - v.0 (2024) 🌎 💻 🌟")
    gr.Markdown("## ✉️✨ Submit your model here!", elem_classes="markdown-text")
    gr.Markdown("## Fill out below then click **Run Analysis** to create the request file and launch the job.")
    gr.Markdown("## The [Project Leaderboard](https://huggingface.co/spaces/EnergyStarAI/2024_Leaderboard) will be updated quarterly, as new models get submitted.")
    with gr.Row():
        with gr.Column():
            task = gr.Dropdown(
                choices=task_mappings.keys(),
                label="Choose a benchmark task",
                value = 'Text Generation',
                multiselect=False,
                interactive=True,
            )
        with gr.Column():
            model_name_textbox = gr.Textbox(label="Model name (user_name/model_name)")

    with gr.Row():
        with gr.Column():
            submit_button = gr.Button("Run Analysis")
            submission_result = gr.Markdown()
            submit_button.click(
                fn=add_new_eval,
                inputs=[
                    model_name_textbox,
                    task,
                ],
                outputs=submission_result,
            )
    with gr.Row():
        with gr.Column():
                with gr.Accordion("Models that are in the latest leaderboard version:", open = False):
                    gr.Dataframe(get_leaderboard_models())
                with gr.Accordion("Models that have been benchmarked lately:", open = False):
                    gr.Dataframe(print_existing_models())
demo.launch()