|
import gradio as gr |
|
import pandas as pd |
|
|
|
from apscheduler.schedulers.background import BackgroundScheduler |
|
|
|
from src.display.css_html_js import custom_css |
|
|
|
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.utils import ( |
|
BENCHMARK_COLS, |
|
COLS, |
|
EVAL_COLS, |
|
EVAL_TYPES, |
|
NUMERIC_INTERVALS, |
|
TYPES, |
|
AutoEvalColumn, |
|
ModelType, |
|
fields, |
|
WeightType, |
|
Precision |
|
) |
|
|
|
from src.populate import get_evaluation_queue_df, get_leaderboard_df |
|
|
|
from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, H4_TOKEN, IS_PUBLIC, QUEUE_REPO, REPO_ID, RESULTS_REPO |
|
from src.submission.submit import add_new_eval |
|
|
|
from src.display.utils import Tasks |
|
|
|
from huggingface_hub import snapshot_download |
|
|
|
|
|
|
|
def restart_space(): |
|
API.restart_space(repo_id=REPO_ID, token=H4_TOKEN) |
|
|
|
def ui_snapshot_download(repo_id, local_dir, repo_type, tqdm_class, etag_timeout): |
|
try: |
|
print(f"local_dir for snapshot download = {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: |
|
print(f"ui_snapshot_download failed. restarting space...") |
|
restart_space() |
|
|
|
|
|
def update_table(hidden_df: pd.DataFrame, columns: list, type_query: list, precision_query: list, size_query: list, query: str): |
|
print(f"hidden_df = {hidden_df}") |
|
show_deleted = True |
|
filtered_df = filter_models(hidden_df, type_query, size_query, precision_query, show_deleted) |
|
|
|
print(f"filtered_df = {filtered_df}") |
|
filtered_df = filter_queries(query, filtered_df) |
|
df = select_columns(filtered_df, columns) |
|
print(f"df = {df}") |
|
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, |
|
] |
|
|
|
filtered_df = df[ |
|
always_here_cols + [c for c in COLS if c in df.columns and c in columns] + [AutoEvalColumn.dummy.name] |
|
] |
|
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) |
|
filtered_df = filtered_df.drop_duplicates( |
|
subset=[AutoEvalColumn.model.name, AutoEvalColumn.precision.name, AutoEvalColumn.revision.name] |
|
) |
|
|
|
return filtered_df |
|
|
|
|
|
def filter_models(df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, show_deleted: bool) -> pd.DataFrame: |
|
|
|
|
|
print("aa this is an example", df) |
|
print(f"filter_models()'s df: {df}\n") |
|
|
|
if show_deleted: |
|
filtered_df = df |
|
else: |
|
filtered_df = df[df[AutoEvalColumn.still_on_hub.name] is True] |
|
|
|
type_emoji = [t[0] for t in type_query] |
|
print("aa this is an example", df, AutoEvalColumn.model_type_symbol.name, "thhhthht") |
|
print("type", type_emoji) |
|
filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)] |
|
print("bb", filtered_df) |
|
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 |
|
|
|
|
|
|
|
|
|
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) |
|
|
|
print(f"COLS = {COLS}") |
|
|
|
|
|
raw_data, original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS) |
|
leaderboard_df = original_df.copy() |
|
print(f"leaderboard_df = {leaderboard_df}") |
|
|
|
|
|
|
|
demo = gr.Blocks(css=custom_css) |
|
with demo: |
|
gr.HTML(TITLE) |
|
gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text") |
|
|
|
with gr.Tabs(elem_classes="tab-buttons") as tabs: |
|
|
|
|
|
with gr.TabItem("Benchmarks", elem_id="llm-benchmark-tab-table", id=0): |
|
|
|
with gr.Row(): |
|
with gr.Column(): |
|
with gr.Row(): |
|
search_bar = gr.Textbox(placeholder=" 🔍 Model search (separate multiple queries with `;`)", show_label=False, elem_id="search-bar",) |
|
with gr.Row(): |
|
shown_columns = gr.CheckboxGroup( |
|
choices=[ |
|
c.name |
|
for c in fields(AutoEvalColumn) |
|
if not c.hidden and not c.never_hidden and not c.dummy |
|
], |
|
value=[ |
|
c.name |
|
for c in fields(AutoEvalColumn) |
|
if c.displayed_by_default and not c.hidden and not c.never_hidden |
|
], |
|
label="Select columns to show", |
|
elem_id="column-select", |
|
interactive=True, |
|
) |
|
|
|
with gr.Column(min_width=320): |
|
filter_columns_type = gr.CheckboxGroup( |
|
label="Model types", |
|
choices=[t.to_str() for t in ModelType], |
|
value=[t.to_str() for t in ModelType], |
|
interactive=True, |
|
elem_id="filter-columns-type", |
|
) |
|
filter_columns_precision = gr.CheckboxGroup( |
|
label="Precision", |
|
choices=[i.value.name for i in Precision], |
|
value=[i.value.name for i in Precision], |
|
interactive=True, |
|
elem_id="filter-columns-precision", |
|
) |
|
filter_columns_size = gr.CheckboxGroup( |
|
label="Model sizes (in billions of parameters)", |
|
choices=list(NUMERIC_INTERVALS.keys()), |
|
value=list(NUMERIC_INTERVALS.keys()), |
|
interactive=True, |
|
elem_id="filter-columns-size", |
|
) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
leaderboard_table = gr.components.Dataframe( |
|
|
|
value=leaderboard_df[ |
|
[c.name for c in fields(AutoEvalColumn) if c.never_hidden] |
|
+ shown_columns.value |
|
+ [AutoEvalColumn.dummy.name] |
|
] if leaderboard_df.empty is False else leaderboard_df, |
|
headers=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + shown_columns.value, |
|
datatype=TYPES, |
|
elem_id="leaderboard-table", |
|
interactive=False, |
|
visible=True, |
|
|
|
) |
|
|
|
hidden_leaderboard_table_for_search = gr.components.Dataframe( |
|
value=original_df[COLS] if original_df.empty is False else original_df, |
|
headers=COLS, |
|
datatype=TYPES, |
|
visible=False |
|
) |
|
for selector in [shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size]: |
|
selector.change( |
|
update_table, |
|
[ |
|
hidden_leaderboard_table_for_search, |
|
shown_columns, |
|
filter_columns_type, |
|
filter_columns_precision, |
|
filter_columns_size, |
|
search_bar, |
|
], |
|
leaderboard_table, |
|
queue=True, |
|
) |
|
|
|
|
|
with gr.TabItem("🚀 Submit here! ", elem_id="llm-benchmark-tab-table", id=3): |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
with gr.Row(): |
|
gr.Markdown("# ✉️✨ Submit your model here!", elem_classes="markdown-text") |
|
|
|
with gr.Row(): |
|
with gr.Column(): |
|
model_name_textbox = gr.Textbox(label="Model name") |
|
|
|
revision_name_textbox = gr.Textbox(label="Revision commit", placeholder="main") |
|
private = gr.Checkbox(False, label="Private", visible=not IS_PUBLIC) |
|
model_type = gr.Dropdown( |
|
choices=[t.to_str(" : ") for t in ModelType if t != ModelType.Unknown], |
|
label="Model type", |
|
multiselect=False, |
|
value=None, |
|
interactive=True, |
|
) |
|
|
|
with gr.Column(): |
|
precision = gr.Dropdown( |
|
choices=[i.value.name for i in Precision if i != Precision.Unknown], |
|
label="Precision", |
|
multiselect=False, |
|
value="float32", |
|
interactive=True, |
|
) |
|
weight_type = gr.Dropdown( |
|
choices=[i.value.name for i in WeightType], |
|
label="Weights type", |
|
multiselect=False, |
|
value="Original", |
|
interactive=True, |
|
) |
|
|
|
|
|
requested_tasks = gr.CheckboxGroup( |
|
choices=[ (i.value.col_name, i.value) for i in Tasks], |
|
|
|
label="Select tasks", |
|
elem_id="task-select", |
|
interactive=True, |
|
) |
|
|
|
|
|
base_model_name_textbox = gr.Textbox(label="Base model (for delta or adapter weights)") |
|
|
|
submit_button = gr.Button("Submit Eval") |
|
submission_result = gr.Markdown() |
|
|
|
|
|
submit_button.click( |
|
add_new_eval, |
|
[ |
|
model_name_textbox, |
|
|
|
requested_tasks, |
|
|
|
base_model_name_textbox, |
|
revision_name_textbox, |
|
precision, |
|
private, |
|
weight_type, |
|
model_type, |
|
], |
|
submission_result) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
scheduler = BackgroundScheduler() |
|
|
|
scheduler.add_job(restart_space, "interval", seconds=6 * 60 * 60) |
|
|
|
scheduler.start() |
|
|
|
|
|
|
|
demo.launch() |