|
import logging |
|
import os |
|
|
|
import gradio as gr |
|
|
|
from src.content import ( |
|
INTRODUCTION_TEXT, |
|
INTRODUCTION_TITLE, |
|
LEADERBOARD_TEXT, |
|
LEADERBOARD_TITLE, |
|
SUBMISSION_TEXT_FILES, |
|
SUBMISSION_TEXT_INTRO, |
|
SUBMISSION_TEXT_METADATA, |
|
SUBMISSION_TEXT_SUBMIT, |
|
SUBMISSION_TEXT_TASK, |
|
SUBMISSION_TITLE, |
|
) |
|
from src.get_results_for_task import get_results_for_task |
|
from src.leaderboard_formatting import get_types_per_task |
|
from src.submission_uploader import SubmissionUploader |
|
from src.tasks_content import ( |
|
TASKS_DESCRIPTIONS, |
|
TASKS_PRETTY, |
|
TASKS_PRETTY_REVERSE, |
|
get_submission_text_files_for_task, |
|
) |
|
|
|
logging.basicConfig( |
|
level=logging.INFO, |
|
format="%(asctime)s [%(levelname)s] %(message)s", |
|
handlers=[logging.StreamHandler()], |
|
) |
|
|
|
submission_uploader = SubmissionUploader( |
|
dataset_id=os.environ["DATASET_ID"], private_dataset_id=os.environ["PRIVATE_DATASET_ID"] |
|
) |
|
|
|
|
|
def get_leaderboard_for_task(task_pretty: str) -> gr.components.Dataframe: |
|
return gr.components.Dataframe( |
|
value=get_results_for_task(task_pretty), |
|
interactive=False, |
|
datatype=get_types_per_task(TASKS_PRETTY_REVERSE[task_pretty]), |
|
) |
|
|
|
code_completion_dataset_names = get_results_for_task(TASKS_PRETTY['project_code_completion'])['Dataset Name'].unique().tolist() |
|
|
|
def get_leaderboard_for_completion_task(dataset_name: str | None): |
|
df = get_results_for_task(TASKS_PRETTY['project_code_completion']) |
|
if dataset_name is None: |
|
dataset_name = code_completion_dataset_names[0] |
|
filtered_df = df[df['Dataset Name']==dataset_name] |
|
filtered_df.drop(columns=['Dataset'], inplace=True) |
|
return gr.components.Dataframe( |
|
value=filtered_df, |
|
interactive=False, |
|
datatype=get_types_per_task('project_code_completion'), |
|
) |
|
|
|
|
|
with gr.Blocks() as demo: |
|
|
|
gr.HTML(INTRODUCTION_TITLE) |
|
gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text") |
|
|
|
|
|
gr.HTML(LEADERBOARD_TITLE) |
|
gr.Markdown(LEADERBOARD_TEXT, elem_classes="markdown-text") |
|
with gr.Tabs(): |
|
for task_pretty in TASKS_PRETTY_REVERSE: |
|
with gr.TabItem(task_pretty): |
|
with gr.Row(): |
|
gr.Markdown(TASKS_DESCRIPTIONS[TASKS_PRETTY_REVERSE[task_pretty]]) |
|
|
|
if task_pretty == TASKS_PRETTY['project_code_completion']: |
|
leaderboard_table = get_leaderboard_for_completion_task(dataset_name=None) |
|
|
|
else: |
|
leaderboard_table = get_leaderboard_for_task(task_pretty) |
|
|
|
task_input = gr.Text(value=task_pretty, visible=False) |
|
|
|
if task_pretty == TASKS_PRETTY['project_code_completion']: |
|
dataset_dropdown = gr.Dropdown(choices=code_completion_dataset_names, label="Select the Dataset") |
|
dataset_dropdown.change( |
|
fn=get_leaderboard_for_completion_task, |
|
inputs=dataset_dropdown, |
|
outputs=leaderboard_table |
|
) |
|
refresh_button = gr.Button("π Refresh", variant="secondary") |
|
refresh_button.click( |
|
fn=get_leaderboard_for_completion_task, |
|
inputs=dataset_dropdown, |
|
outputs=leaderboard_table, |
|
) |
|
else: |
|
refresh_button = gr.Button("π Refresh", variant="secondary") |
|
refresh_button.click( |
|
fn=get_leaderboard_for_task, |
|
inputs=task_input, |
|
outputs=leaderboard_table, |
|
) |
|
|
|
|
|
gr.HTML(SUBMISSION_TITLE) |
|
gr.Markdown(SUBMISSION_TEXT_INTRO, elem_classes="markdown-text") |
|
with gr.Accordion("π Submit new results", open=False): |
|
gr.Markdown(SUBMISSION_TEXT_TASK, elem_classes="markdown-text") |
|
task_selection = gr.Radio(TASKS_PRETTY_REVERSE.keys(), label="Task") |
|
|
|
gr.Markdown(SUBMISSION_TEXT_METADATA, elem_classes="markdown-text") |
|
with gr.Row(): |
|
with gr.Column(): |
|
model_folder_textbox = gr.Textbox( |
|
label="Model Folder", |
|
placeholder="How to call a folder related to this submission in our results dataset (should be unique).", |
|
) |
|
model_name_textbox = gr.Textbox( |
|
label="Model Name", |
|
placeholder="How to display model's name on the leaderboard.", |
|
) |
|
model_url_textbox = gr.Textbox( |
|
label="Model URL", |
|
placeholder="Link to a model's page - will be clickable on a leaderboard (optional).", |
|
) |
|
|
|
with gr.Column(): |
|
url_textbox = gr.Textbox( |
|
label="Relevant URLs", |
|
placeholder='URLs to relevant resources with additional details about your submission (optional). Use the following format: "[text1](link1), [text2](link2)".', |
|
) |
|
model_availability_textbox = gr.Textbox( |
|
label="Availability", |
|
placeholder="Information about the model's availability and licensing.", |
|
) |
|
context_size_textbox = gr.Textbox( |
|
label="Context Size", |
|
placeholder="Context size in tokens used for the submission (should be an integer).", |
|
) |
|
with gr.Column(): |
|
submitted_by_textbox = gr.Textbox( |
|
label="Submitted By", |
|
placeholder="How to display on the leaderboard who submitted the model.", |
|
) |
|
contact_textbox = gr.Textbox( |
|
label="Contact Information", |
|
placeholder="How Long Code Arena team can contact you (won't go to public dataset).", |
|
) |
|
comment_textbox = gr.Textbox( |
|
label="Comment", |
|
placeholder="Any comments you have for Long Code Arena team (optional, won't go to public dataset).", |
|
) |
|
|
|
gr.Markdown(SUBMISSION_TEXT_FILES, elem_classes="markdown-text") |
|
with gr.Row(): |
|
with gr.Column(variant="panel"): |
|
task_specific_instructions = gr.Markdown(get_submission_text_files_for_task(None)) |
|
task_selection.select(get_submission_text_files_for_task, [task_selection], task_specific_instructions) |
|
with gr.Column(): |
|
file_output = gr.File(file_count="multiple") |
|
|
|
gr.Markdown(SUBMISSION_TEXT_SUBMIT, elem_classes="markdown-text") |
|
submit_button = gr.Button("Submit") |
|
submission_result = gr.Markdown() |
|
submit_button.click( |
|
submission_uploader.upload_files, |
|
[ |
|
task_selection, |
|
model_folder_textbox, |
|
model_name_textbox, |
|
model_availability_textbox, |
|
model_url_textbox, |
|
url_textbox, |
|
context_size_textbox, |
|
submitted_by_textbox, |
|
contact_textbox, |
|
comment_textbox, |
|
file_output, |
|
], |
|
submission_result, |
|
) |
|
|
|
if __name__ == "__main__": |
|
demo.queue() |
|
demo.launch() |
|
|