utils.py

import streamlit as st
import pandas as pd
import wandb
import time
from datetime import datetime, timedelta
import requests


def get_competitions(CONFIG_URL):
 """
 Get competition names and their evaluation times from the config URL.
 """
 competitions = []
 try:
 config = requests.get(CONFIG_URL).json()
 for competition in config:
 # Extract competition name and evaluation times
 competition_name = competition["competition_id"]
 evaluation_times = competition.get("evaluation_times", [])
 
 # Store competition name and evaluation times as a tuple
 competitions.append((competition_name, evaluation_times))
 except Exception as e:
 print(f"Error loading competition data: {str(e)}")
 
 return competitions


def get_latest_evaluation_time(evaluation_times):
 """
 Get the latest UTC evaluation time as a datetime object (either today or yesterday).
 
 Args:
 evaluation_times (list): List of evaluation times as strings (in HH:MM format).
 
 Returns:
 datetime: Latest evaluation time as a datetime object.
 """
 
 # Get the current UTC date and time
 current_utc_datetime = datetime.utcnow()
 current_utc_time = current_utc_datetime.time()
 
 # Convert evaluation times to datetime.time objects
 eval_times = [datetime.strptime(time, "%H:%M").time() for time in evaluation_times]
 
 # Sort the evaluation times to ensure they are in chronological order
 eval_times.sort()
 
 # Loop through the evaluation times in reverse to find the latest time that has passed
 for eval_time in reversed(eval_times):
 if current_utc_time >= eval_time:
 # If the time has passed today, return today's date with that time
 return current_utc_datetime.replace(hour=eval_time.hour, minute=eval_time.minute, second=0, microsecond=0)
 
 # If none of the evaluation times have passed, return the latest one from the previous day
 yesterday = current_utc_datetime - timedelta(days=1)
 latest_eval_time = eval_times[-1]
 
 return yesterday.replace(hour=latest_eval_time.hour, minute=latest_eval_time.minute, second=0, microsecond=0)


def fetch_competition_summary(api, entity, project):
 data = []
 
 # entity = projects[selected_project]["entity"]
 # project = projects[selected_project]["project"]
 
 runs = api.runs(f"{entity}/{project}")
 
 for run in runs:
 try:
 summary = run.summary
 if summary.get("validator_hotkey") and summary.get("winning_hotkey"):
 data.append({
 "Created At": run.created_at,
 "Validator ID": summary.get("validator_hotkey"),
 "Winning Hotkey": summary.get("winning_hotkey"),
 "Run Time (s)": summary.get("run_time_s"),
 })
 except Exception as e:
 st.write(f"Error processing run {run.id}: {str(e)}")
 
 df = pd.DataFrame(data)
 if not df.empty:
 df['Created At'] = pd.to_datetime(df['Created At'], utc=True)
 df = df.sort_values(by="Created At", ascending=False)
 
 return df

def fetch_models_evaluation(api, entity, project):
 data = []
 
 # entity = projects[selected_project]["entity"]
 # project = projects[selected_project]["project"]
 
 runs = api.runs(f"{entity}/{project}")
 
 for run in runs:
 try:
 summary = run.summary
 if summary.get("score") is not None: # Assuming runs with score are model evaluations
 model_link = summary.get("hf_model_link", "")
 if model_link:
 # Create clickable link
 model_link_html = f'<a href="{model_link}" target="_blank">Model Link</a>'
 else:
 model_link_html = "N/A"

 data.append({
 "Created At": run.created_at,

 "Miner hotkey": summary.get("miner_hotkey", "N/A"),

 "F1-beta": summary.get("fbeta"),
 "Accuracy": summary.get("accuracy"),
 "Recall": summary.get("recall"),
 "Precision": summary.get("precision"),

 "Tested entries": summary.get("tested_entries"),

 "ROC AUC": summary.get("roc_auc"),
 "Confusion Matrix": summary.get("confusion_matrix"),

 "Model link": model_link_html,
 "Score": summary.get("score"),
 #TODO link to huggingface model
 })
 except Exception as e:
 st.write(f"Error processing run {run.id}: {str(e)}")
 
 df = pd.DataFrame(data)
 if not df.empty:
 df['Created At'] = pd.to_datetime(df['Created At'], utc=True)
 df = df.sort_values(by="Created At", ascending=False)
 
 return df

def highlight_score_column(s):
 """
 Highlight the 'Score' column with a custom background color.
 """
 return ['background-color: yellow' if s.name == 'Score' else '' for _ in s]