Spaces:
Running
Running
| import pandas as pd | |
| from pathlib import Path | |
| from ..styles import highlight_color | |
| abs_path = Path(__file__).parent.parent.parent | |
| def replace_models_names(model_name): | |
| if "gpt" in model_name: | |
| return model_name | |
| replaces = {'meta-llama': 'meta_llama', | |
| 'epfl-llm':'epfl_llm', | |
| '01-ai':'01_ai'} | |
| new_name = model_name.replace('model-', '') | |
| for k, v in replaces.items(): | |
| if new_name.startswith(k): | |
| new_name = new_name.replace(k, v) | |
| new_name = new_name.replace('-','/',1) | |
| new_name = new_name.replace('_','-',1) | |
| new_name = f"[{new_name}](https://huggingface.co/{new_name})" | |
| return new_name | |
| def generate_ORDER_LIST_LDEK_and_data_types(json_data): | |
| ORDER_LIST_LDEK = ["model_name", "overall_accuracy"] | |
| data_types = ["markdown", "number"] | |
| for key in json_data.keys(): | |
| if key not in ["model_name", "overall_accuracy"]: | |
| ORDER_LIST_LDEK.append(key) | |
| data_types.append("number") | |
| ORDER_LIST_LDEK[2:] = sorted(ORDER_LIST_LDEK[2:]) | |
| return ORDER_LIST_LDEK, data_types | |
| def filter_columns_ldek(column_choices): | |
| selected_columns = [col for col in ORDER_LIST_LDEK if col in column_choices] | |
| return LDEK_ACCS[selected_columns] | |
| def load_json_data(file_path, ORDER_LIST_LDEK): | |
| LDEK_ACCS = pd.read_json(file_path) | |
| for column in LDEK_ACCS.columns: | |
| if LDEK_ACCS[column].apply(type).eq(dict).any(): | |
| LDEK_ACCS[column] = LDEK_ACCS[column].apply(str) | |
| LDEK_ACCS["model_name"] = LDEK_ACCS["model_name"].apply( | |
| lambda name: replace_models_names(name) | |
| ) | |
| for column in LDEK_ACCS.select_dtypes(include='number').columns: | |
| LDEK_ACCS[column] = LDEK_ACCS[column].round(2) | |
| LDEK_ACCS["overall_accuracy"] = pd.to_numeric(LDEK_ACCS["overall_accuracy"], errors='coerce') | |
| ordered_columns = [col for col in ORDER_LIST_LDEK if col in LDEK_ACCS.columns] | |
| LDEK_ACCS = LDEK_ACCS[ordered_columns] | |
| if "overall_accuracy" in LDEK_ACCS.columns: | |
| LDEK_ACCS = LDEK_ACCS.sort_values(by="overall_accuracy", ascending=False) | |
| return LDEK_ACCS | |
| file_path = str(abs_path / "leaderboards/ldek_accs.json") | |
| with open(file_path, 'r', encoding='utf-8') as file: | |
| sample_data = pd.read_json(file_path).iloc[0].to_dict() | |
| ORDER_LIST_LDEK, DATA_TYPES_LDEK = generate_ORDER_LIST_LDEK_and_data_types(sample_data) | |
| LDEK_ACCS = load_json_data(file_path, ORDER_LIST_LDEK) | |
| LDEK_ACCS = LDEK_ACCS.style.highlight_max( | |
| color = highlight_color, | |
| subset=LDEK_ACCS.columns[1:]).format(precision=2) | |
| COLUMN_HEADERS_LDEK = ORDER_LIST_LDEK | |
| print(ORDER_LIST_LDEK) | |
| file_path2 = str(abs_path / "leaderboards/ldek_en_accs.json") | |
| with open(file_path, 'r', encoding='utf-8') as file: | |
| sample_data2 = pd.read_json(file_path).iloc[0].to_dict() | |
| ORDER_LIST_LDEK_EN, DATA_TYPES_LDEK_EN = generate_ORDER_LIST_LDEK_and_data_types(sample_data2) | |
| LDEK_EN_ACCS = load_json_data(file_path2, ORDER_LIST_LDEK_EN) | |
| LDEK_EN_ACCS = LDEK_EN_ACCS.style.highlight_max( | |
| color = highlight_color, | |
| subset=LDEK_EN_ACCS.columns[1:]).format(precision=2) | |
| COLUMN_HEADERS_LDEK_EN = ORDER_LIST_LDEK_EN | |
| print(ORDER_LIST_LDEK_EN) | |