app.py

#!/usr/bin/env python

import gradio as gr
import pandas as pd

from apscheduler.schedulers.background import BackgroundScheduler

from huggingface_hub import snapshot_download

from src.display.about import (
    CITATION_BUTTON_LABEL,
    CITATION_BUTTON_TEXT,
    EVALUATION_QUEUE_TEXT,
    INTRODUCTION_TEXT,
    LLM_BENCHMARKS_TEXT,
    LLM_BENCHMARKS_DETAILS,
    FAQ_TEXT,
    TITLE
)

from src.display.css_html_js import custom_css

from src.display.utils import (
    BENCHMARK_COLS,
    COLS,
    EVAL_COLS,
    EVAL_TYPES,
    NUMERIC_INTERVALS,
    TYPES,
    AutoEvalColumn,
    ModelType,
    fields,
    WeightType,
    Precision
)

from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, H4_TOKEN, IS_PUBLIC, QUEUE_REPO, REPO_ID, RESULTS_REPO
from src.populate import get_evaluation_queue_df, get_leaderboard_df
from src.submission.submit import add_new_eval
from src.utils import get_dataset_summary_table


def ui_snapshot_download(repo_id, local_dir, repo_type, tqdm_class, etag_timeout):
    try:
        print(local_dir)
        snapshot_download(repo_id=repo_id, local_dir=local_dir, repo_type=repo_type, tqdm_class=tqdm_class, etag_timeout=etag_timeout)
    except Exception as e:
        restart_space()


def restart_space():
    API.restart_space(repo_id=REPO_ID, token=H4_TOKEN)


def init_space():
    dataset_df = get_dataset_summary_table(file_path='blog/Hallucination-Leaderboard-Summary.csv')

    import socket
    if socket.gethostname() not in {'neuromancer'}:
        ui_snapshot_download(repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30)
        ui_snapshot_download(repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30)

    raw_data, original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)

    finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS)
    return dataset_df, original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df


dataset_df, original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = init_space()
leaderboard_df = original_df.copy()


# Searching and filtering
def update_table(hidden_df: pd.DataFrame,
                 columns: list,
                 type_query: list,
                 precision_query: list,
                 size_query: list,
                 query: str):
    filtered_df = filter_models(hidden_df, type_query, size_query, precision_query)
    filtered_df = filter_queries(query, filtered_df)
    df = select_columns(filtered_df, columns)
    return df


def search_table(df: pd.DataFrame, query: str) -> pd.DataFrame:
    return df[(df[AutoEvalColumn.dummy.name].str.contains(query, case=False))]


def select_columns(df: pd.DataFrame, columns: list) -> pd.DataFrame:
    # always_here_cols = [AutoEvalColumn.model_type_symbol.name, AutoEvalColumn.model.name]

    always_here_cols = [c.name for c in fields(AutoEvalColumn) if c.never_hidden]
    dummy_col = [AutoEvalColumn.dummy.name]

    # We use COLS to maintain sorting
    filtered_df = df[
        # always_here_cols + [c for c in COLS if c in df.columns and c in columns] + [AutoEvalColumn.dummy.name]
        always_here_cols + [c for c in COLS if c in df.columns and c in columns] + dummy_col
    ]
    return filtered_df


def filter_queries(query: str, filtered_df: pd.DataFrame):
    final_df = []
    if query != "":
        queries = [q.strip() for q in query.split(";")]
        for _q in queries:
            _q = _q.strip()
            if _q != "":
                temp_filtered_df = search_table(filtered_df, _q)
                if len(temp_filtered_df) > 0:
                    final_df.append(temp_filtered_df)
        if len(final_df) > 0:
            filtered_df = pd.concat(final_df)
            subset = [AutoEvalColumn.model.name, AutoEvalColumn.precision.name, AutoEvalColumn.revision.name]
            filtered_df = filtered_df.drop_duplicates(subset=subset)
    return filtered_df


def filter_models(df: pd.DataFrame,
                  type_query: list,
                  size_query: list,
                  precision_query: list) -> pd.DataFrame:
    # Show all models
    filtered_df = df

    type_emoji = [t[0] for t in type_query]
    filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
    filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isin(precision_query + ["None"])]

    numeric_interval = pd.IntervalIndex(sorted([NUMERIC_INTERVALS[s] for s in size_query]))
    params_column = pd.to_numeric(df[AutoEvalColumn.params.name], errors="coerce")
    mask = params_column.apply(lambda x: any(numeric_interval.contains(x)))
    filtered_df = filtered_df.loc[mask]

    return filtered_df


# triggered only once at startup => read query parameter if it exists
def load_query(request: gr.Request):
    query = request.query_params.get("query") or ""
    return query

leaderboard_df = filter_models(
    df=leaderboard_df,
    type_query=[t.to_str(" : ") for t in ModelType],
    size_query=list(NUMERIC_INTERVALS.keys()),
    precision_query=[i.value.name for i in Precision],
)

import unicodedata

def is_valid_unicode(char):
    try:
        unicodedata.name(char)
        return True  # Valid Unicode character
    except ValueError:
        return False  # Invalid Unicode character

def remove_invalid_unicode(input_string):
    if isinstance(input_string, str):
        valid_chars = [char for char in input_string if is_valid_unicode(char)]
        return ''.join(valid_chars)
    else:
        return input_string  # Return non-string values as is

dummy1 = gr.Textbox(visible=False)

hidden_leaderboard_table_for_search = gr.components.Dataframe(
    headers=COLS,
    datatype=TYPES,
    visible=False,
    line_breaks=False,
    interactive=False
)

def display(x, y):
    # Assuming df is your DataFrame
    for column in leaderboard_df.columns:
        if leaderboard_df[column].dtype == 'object':
            leaderboard_df[column] = leaderboard_df[column].apply(remove_invalid_unicode)

    subset_df = leaderboard_df[COLS]
    return subset_df

INTRODUCTION_TEXT = """
This is a copied space from LLM Trustworthy Leaderboard. Instead of displaying
the results as table this space was modified to simply provides a gradio API interface. 
Using the following python script below, users can access the full leaderboard data easily.
Python on how to access the data:
```python
# Import dependencies
from gradio_client import Client
# Initialize the Gradio client with the API URL
client = Client("https://rodrigomasini-data-only-hallucination-leaderboard.hf.space/")
try:
    # Perform the API call
    response = client.predict("","", api_name='/predict')
    # Check if response it's directly accessible
    if len(response) > 0:
        print("Response received!")
        headers = response.get('headers', [])
        data = response.get('data', [])
        print(headers)
        # Remove commenst if you want to download the dataset and save in csv format
        # Specify the path to your CSV file
        #csv_file_path = 'llm-trustworthy-benchmark.csv'
        # Open the CSV file for writing
        #with open(csv_file_path, mode='w', newline='', encoding='utf-8') as file:
        #    writer = csv.writer(file)
            # Write the headers
        #    writer.writerow(headers)
            # Write the data
        #    for row in data:
        #        writer.writerow(row)
        #print(f"Results saved to {csv_file_path}")
    # If the above line prints a string that looks like JSON, you can parse it with json.loads(response)
    # Otherwise, you might need to adjust based on the actual structure of `response`
except Exception as e:
    print(f"An error occurred: {e}")
```
"""

interface = gr.Interface(
    fn=display,
    inputs=[gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text"), dummy1],
    outputs=[hidden_leaderboard_table_for_search]
)

scheduler = BackgroundScheduler()
scheduler.add_job(restart_space, "interval", seconds=1800)

scheduler.start()

interface.launch()