working generic evaluation

.dockerignore

@@ -0,0 +1,146 @@
+# Local stuff
+.DS_Store
+.vscode/
+.vim/
+flagged/
+*.csv
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+pip-wheel-metadata/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+.python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+*.env
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# Terraform stuff
+*.tfstate
+*.tfstate.backup
+.terraform**
+**.tfvars
+
+# Alembic / database artifcats
+**.db

Dockerfile

@@ -1,9 +1,9 @@
-FROM python:3.8.9
+FROM ubuntu:22.04
 
 ENV DEBIAN_FRONTEND=noninteractive \
     TZ=UTC
 
-RUN pip install pip==23.0.1
+RUN pip install pip==23.3.2
 
 WORKDIR /app
 RUN mkdir -p /app/.cache
@@ -14,10 +14,6 @@ ENV HOME=/app
 
 ENV PYTHONPATH=$HOME/app \
     PYTHONUNBUFFERED=1 \
-    GRADIO_ALLOW_FLAGGING=never \
-    GRADIO_NUM_PORTS=1 \
-    GRADIO_SERVER_NAME=0.0.0.0 \
-    GRADIO_THEME=huggingface \
     SYSTEM=spaces
 
 
@@ -26,7 +22,8 @@ RUN wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh \
     && rm -f Miniconda3-latest-Linux-x86_64.sh
 ENV PATH /app/miniconda/bin:$PATH
 
-RUN conda create -p /app/env -y python=3.8
+RUN conda create -p /app/env -y python=3.10 \
+    && conda clean -ya
 
 
 SHELL ["conda", "run","--no-capture-output", "-p","/app/env", "/bin/bash", "-c"]

competitions/__init__.py

@@ -1,7 +1,5 @@
 import os
 
-from .info import CompetitionInfo
-
 
 __version__ = "0.1.1"
 
@@ -9,10 +7,3 @@ MOONLANDING_URL = os.getenv("MOONLANDING_URL", "https://huggingface.co")
 COMPETITION_ID = os.getenv("COMPETITION_ID")
 AUTOTRAIN_USERNAME = os.getenv("AUTOTRAIN_USERNAME")
 AUTOTRAIN_TOKEN = os.getenv("AUTOTRAIN_TOKEN")
-AUTOTRAIN_BACKEND_API = os.getenv("AUTOTRAIN_BACKEND_API", "https://api.autotrain.huggingface.co")
-BOT_TOKEN = os.getenv("BOT_TOKEN")
-
-if COMPETITION_ID is not None:
-    competition_info = CompetitionInfo(competition_id=COMPETITION_ID, autotrain_token=AUTOTRAIN_TOKEN)
-else:
-    competition_info = None

competitions/app.py

@@ -1,4 +1,5 @@
 import os
+import threading
 
 from fastapi import FastAPI, File, Form, Request, UploadFile
 from fastapi.responses import HTMLResponse, JSONResponse
@@ -8,12 +9,14 @@ from pydantic import BaseModel
 
 from competitions.info import CompetitionInfo
 from competitions.leaderboard import Leaderboard
+from competitions.runner import JobRunner
 from competitions.submissions import Submissions
 
 
 HF_TOKEN = os.environ.get("HF_TOKEN", None)
 BASE_DIR = os.path.dirname(os.path.abspath(__file__))
 COMPETITION_ID = os.getenv("COMPETITION_ID")
+OUTPUT_PATH = os.getenv("OUTPUT_PATH", "/tmp/model")
 COMP_INFO = CompetitionInfo(competition_id=COMPETITION_ID, autotrain_token=HF_TOKEN)
 
 
@@ -21,6 +24,15 @@ class User(BaseModel):
     user_token: str
 
 
+def run_job_runner():
+    job_runner = JobRunner(token=HF_TOKEN, competition_info=COMP_INFO, output_path=OUTPUT_PATH)
+    job_runner.run()
+
+
+thread = threading.Thread(target=run_job_runner)
+thread.start()
+
+
 app = FastAPI()
 static_path = os.path.join(BASE_DIR, "static")
 app.mount("/static", StaticFiles(directory=static_path), name="static")
@@ -107,9 +119,9 @@ async def new_submission(
         token=HF_TOKEN,
     )
     if COMP_INFO.competition_type == "generic":
-        resp = sub.new_submission(token, submission_file)
+        resp = sub.new_submission(token, submission_file, submission_comment)
         return {"response": f"Success! You have {resp} submissions remaining today."}
     elif COMP_INFO.competition_type == "code":
-        resp = sub.new_submission(token, hub_model)
+        resp = sub.new_submission(token, hub_model, submission_comment)
         return {"response": f"Success! You have {resp} submissions remaining today."}
     return {"response": "Invalid competition type"}

competitions/competitions.py

@@ -1,184 +0,0 @@
-from datetime import datetime
-from functools import partial
-
-import gradio as gr
-
-from . import AUTOTRAIN_BACKEND_API, AUTOTRAIN_TOKEN, AUTOTRAIN_USERNAME, COMPETITION_ID, competition_info
-from .errors import PastDeadlineError, SubmissionError, SubmissionLimitError
-from .leaderboard import Leaderboard
-from .submissions import Submissions
-from .text import (
-    NO_SUBMISSIONS,
-    SUBMISSION_LIMIT_REACHED,
-    SUBMISSION_SELECTION_TEXT,
-    SUBMISSION_SUCCESS,
-    SUBMISSION_TEXT,
-)
-
-
-leaderboard = Leaderboard(
-    end_date=competition_info.end_date,
-    eval_higher_is_better=competition_info.eval_higher_is_better,
-    max_selected_submissions=competition_info.selection_limit,
-    competition_id=COMPETITION_ID,
-    autotrain_token=AUTOTRAIN_TOKEN,
-)
-
-submissions = Submissions(
-    competition_id=competition_info.competition_id,
-    submission_limit=competition_info.submission_limit,
-    end_date=competition_info.end_date,
-    autotrain_username=AUTOTRAIN_USERNAME,
-    autotrain_token=AUTOTRAIN_TOKEN,
-    autotrain_backend_api=AUTOTRAIN_BACKEND_API,
-)
-
-
-def _new_submission(user_token, submission_file):
-    try:
-        remaining_subs = submissions.new_submission(user_token, submission_file)
-        return SUBMISSION_SUCCESS.format(remaining_subs)
-    except SubmissionLimitError:
-        return SUBMISSION_LIMIT_REACHED
-    except SubmissionError:
-        return "Something went wrong. Please try again later."
-
-
-def _my_submissions(user_token):
-    df, failed_df = submissions.my_submissions(user_token)
-    if len(df) == 0:
-        return [
-            gr.Markdown.update(visible=True, value=NO_SUBMISSIONS),
-            gr.DataFrame.update(visible=False),
-            gr.DataFrame.update(
-                visible=True if len(failed_df) > 0 else False, value=failed_df if len(failed_df) > 0 else None
-            ),
-            gr.TextArea.update(visible=False),
-            gr.Button.update(visible=False),
-        ]
-    selected_submission_ids = df[df["selected"] == True]["submission_id"].values.tolist()
-    failed_selected_submission_ids = failed_df[failed_df["selected"] == True]["submission_id"].values.tolist()
-    selected_submission_ids.extend(failed_selected_submission_ids)
-    if len(selected_submission_ids) > 0:
-        return [
-            gr.Markdown.update(visible=True),
-            gr.DataFrame.update(visible=True, value=df),
-            gr.DataFrame.update(
-                visible=True if len(failed_df) > 0 else False, value=failed_df if len(failed_df) > 0 else None
-            ),
-            gr.TextArea.update(visible=True, value="\n".join(selected_submission_ids), interactive=True),
-            gr.Button.update(visible=True),
-        ]
-    return [
-        gr.Markdown.update(visible=False),
-        gr.DataFrame.update(visible=True, value=df),
-        gr.DataFrame.update(
-            visible=True if len(failed_df) > 0 else False, value=failed_df if len(failed_df) > 0 else None
-        ),
-        gr.TextArea.update(visible=True, interactive=True),
-        gr.Button.update(visible=True),
-    ]
-
-
-def _update_selected_submissions(user_token, submission_ids):
-    submission_ids = submission_ids.split("\n")
-    submission_ids = [sid.strip() for sid in submission_ids]
-    submission_ids = [sid for sid in submission_ids if len(sid) > 0]
-    if len(submission_ids) > competition_info.selection_limit:
-        raise ValueError(
-            f"You can select only {competition_info.selection_limit} submissions. You selected {len(submission_ids)} submissions."
-        )
-    try:
-        submissions.update_selected_submissions(user_token, submission_ids)
-    except PastDeadlineError:
-        return [
-            gr.Markdown.update(visible=True, value="You can no longer select submissions after the deadline."),
-            gr.DataFrame.update(visible=False),
-            gr.DataFrame.update(visible=False),
-            gr.TextArea.update(visible=False),
-            gr.Button.update(visible=False),
-        ]
-    return _my_submissions(user_token)
-
-
-def _fetch_leaderboard(private):
-    if private:
-        current_date_time = datetime.now()
-        if current_date_time < competition_info.end_date:
-            return [
-                gr.DataFrame.update(visible=False),
-                gr.Markdown.update(
-                    visible=True, value=f"Private Leaderboard will be available on {competition_info.end_date} UTC."
-                ),
-            ]
-    df = leaderboard.fetch(private=private)
-    # df["name"] = df["name"].apply(make_clickable_user)
-    # df.to_csv("public_leaderboard.csv" if not private else "private_leaderboard.csv", index=False)
-    num_teams = len(df)
-    return [
-        gr.DataFrame.update(visible=True, value=df),
-        gr.Markdown.update(visible=True, value=f"Number of teams: {num_teams}"),
-    ]
-
-
-with gr.Blocks(css=".tabitem {padding: 25px}") as demo:
-    with gr.Tabs() as tab_container:
-        with gr.TabItem("Overview", id="overview"):
-            gr.Markdown(f"{competition_info.competition_description}")
-        with gr.TabItem("Dataset", id="dataset_tab") as dataset_tab:
-            gr.Markdown(f"{competition_info.dataset_description}")
-        with gr.TabItem("Public Leaderboard", id="public_leaderboard") as public_leaderboard:
-            output_text_public = gr.Markdown()
-            output_df_public = gr.DataFrame(row_count=(50, "dynamic"), visible=False)
-        with gr.TabItem("Private Leaderboard", id="private_leaderboard") as private_leaderboard:
-            output_text_private = gr.Markdown()
-            output_df_private = gr.DataFrame(row_count=(50, "dynamic"), visible=False)
-        with gr.TabItem("New Submission", id="new_submission"):
-            if competition_info.submission_desc is None:
-                gr.Markdown(SUBMISSION_TEXT.format(competition_info.submission_limit))
-            else:
-                gr.Markdown(f"{competition_info.submission_desc}")
-            user_token = gr.Textbox(
-                max_lines=1, value="", label="Please enter your Hugging Face token (read only)", type="password"
-            )
-            uploaded_file = gr.File()
-            output_text = gr.Markdown(visible=True, show_label=False)
-            new_sub_button = gr.Button("Upload Submission")
-            new_sub_button.click(
-                fn=_new_submission,
-                inputs=[user_token, uploaded_file],
-                outputs=[output_text],
-            )
-        with gr.TabItem("My Submissions", id="my_submissions"):
-            gr.Markdown(SUBMISSION_SELECTION_TEXT.format(competition_info.selection_limit))
-            user_token = gr.Textbox(
-                max_lines=1, value="", label="Please enter your Hugging Face token (read only)", type="password"
-            )
-            output_text = gr.Markdown(visible=True, show_label=False)
-            output_df = gr.DataFrame(visible=False, label="Succesful Submissions")
-            failed_df = gr.DataFrame(visible=False, label="Failed Submissions")
-            selected_submissions = gr.TextArea(
-                visible=False,
-                label="Selected Submissions (one submission id per line)",
-                max_lines=competition_info.selection_limit,
-                lines=competition_info.selection_limit,
-            )
-            update_selected_submissions = gr.Button("Update Selected Submissions", visible=False)
-            my_subs_button = gr.Button("Fetch Submissions")
-            my_subs_button.click(
-                fn=_my_submissions,
-                inputs=[user_token],
-                outputs=[output_text, output_df, failed_df, selected_submissions, update_selected_submissions],
-            )
-            update_selected_submissions.click(
-                fn=_update_selected_submissions,
-                inputs=[user_token, selected_submissions],
-                outputs=[output_text, output_df, failed_df, selected_submissions, update_selected_submissions],
-            )
-
-        fetch_lb_partial = partial(_fetch_leaderboard, private=False)
-        public_leaderboard.select(fetch_lb_partial, inputs=[], outputs=[output_df_public, output_text_public])
-        fetch_lb_partial_private = partial(_fetch_leaderboard, private=True)
-        private_leaderboard.select(
-            fetch_lb_partial_private, inputs=[], outputs=[output_df_private, output_text_private]
-        )

competitions/compute_metrics.py

@@ -0,0 +1,58 @@
+from functools import partial
+
+import pandas as pd
+from huggingface_hub import hf_hub_download
+from sklearn import metrics
+
+
+def compute_metrics(params):
+    solution_file = hf_hub_download(
+        repo_id=params.competition_id,
+        filename="solution.csv",
+        token=params.token,
+        repo_type="dataset",
+    )
+
+    solution_df = pd.read_csv(solution_file)
+
+    submission_filename = f"submissions/{params.user_id}-{params.submission_id}.csv"
+    submission_file = hf_hub_download(
+        repo_id=params.competition_id,
+        filename=submission_filename,
+        token=params.token,
+        repo_type="dataset",
+    )
+    submission_df = pd.read_csv(submission_file)
+
+    public_ids = solution_df[solution_df.split == "public"][params.submission_id_col].values
+    private_ids = solution_df[solution_df.split == "private"][params.submission_id_col].values
+
+    public_solution_df = solution_df[solution_df[params.submission_id_col].isin(public_ids)]
+    public_submission_df = submission_df[submission_df[params.submission_id_col].isin(public_ids)]
+
+    private_solution_df = solution_df[solution_df[params.submission_id_col].isin(private_ids)]
+    private_submission_df = submission_df[submission_df[params.submission_id_col].isin(private_ids)]
+
+    public_solution_df = public_solution_df.sort_values(params.submission_id_col).reset_index(drop=True)
+    public_submission_df = public_submission_df.sort_values(params.submission_id_col).reset_index(drop=True)
+
+    private_solution_df = private_solution_df.sort_values(params.submission_id_col).reset_index(drop=True)
+    private_submission_df = private_submission_df.sort_values(params.submission_id_col).reset_index(drop=True)
+
+    if params.metric == "f1-macro":
+        _metric = partial(metrics.f1_score, average="macro")
+        target_cols = [col for col in solution_df.columns if col not in [params.submission_id_col, "split"]]
+        public_score = _metric(public_solution_df[target_cols], public_submission_df[target_cols])
+        private_score = _metric(private_solution_df[target_cols], private_submission_df[target_cols])
+    else:
+        _metric = getattr(metrics, params.metric)
+        target_cols = [col for col in solution_df.columns if col not in [params.submission_id_col, "split"]]
+        public_score = _metric(private_solution_df[target_cols], public_submission_df[target_cols])
+        private_score = _metric(private_solution_df[target_cols], private_submission_df[target_cols])
+
+    # scores can also be dictionaries for multiple metrics
+    evaluation = {
+        "public_score": public_score,
+        "private_score": private_score,
+    }
+    return evaluation

competitions/evaluate.py

@@ -0,0 +1,49 @@
+import argparse
+import json
+
+from huggingface_hub import snapshot_download
+from loguru import logger
+
+from competitions import utils
+from competitions.compute_metrics import compute_metrics
+from competitions.params import EvalParams
+
+
+def parse_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--config", type=str, required=True)
+    return parser.parse_args()
+
+
+def generate_submission_file(params):
+    logger.info("Downloading submission dataset")
+    snapshot_download(
+        repo_id=params.data_path,
+        local_dir=params.output_path,
+        token=params.token,
+        repo_type="dataset",
+    )
+
+
+@utils.monitor
+def run(params):
+    if isinstance(params, dict):
+        params = EvalParams(**params)
+
+    utils.update_submission_status(params, "processing")
+
+    if params.competition_type == "code":
+        generate_submission_file(params)
+
+    public_score, private_score = compute_metrics(params)
+
+    utils.update_submission_score(params, public_score, private_score)
+    utils.update_submission_status(params, "success")
+    utils.pause_space(params)
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    _params = json.load(open(args.config, encoding="utf-8"))
+    _params = EvalParams(**_params)
+    run(_params)

competitions/info.py

@@ -107,3 +107,21 @@ class CompetitionInfo:
     @property
     def competition_type(self):
         return self.config["COMPETITION_TYPE"].lower().strip()
+
+    @property
+    def metric(self):
+        return self.config["EVAL_METRIC"]
+
+    @property
+    def submission_id_col(self):
+        return self.config["SUBMISSION_ID_COLUMN"]
+
+    @property
+    def submission_cols(self):
+        cols = self.config["SUBMISSION_COLUMNS"].split(",")
+        cols = [c.strip() for c in cols]
+        return cols
+
+    @property
+    def submission_rows(self):
+        return self.config["SUBMISSION_ROWS"]

competitions/params.py

@@ -0,0 +1,30 @@
+import os
+from typing import List
+
+from pydantic import BaseModel
+
+
+class EvalParams(BaseModel):
+    competition_id: str
+    competition_type: str
+    metric: str
+    token: str
+    user_id: str
+    submission_id: str
+    submission_id_col: str
+    submission_cols: List[str]
+    submission_rows: int
+    output_path: str
+
+    class Config:
+        protected_namespaces = ()
+
+    def save(self, output_dir):
+        """
+        Save parameters to a json file.
+        """
+        os.makedirs(output_dir, exist_ok=True)
+        path = os.path.join(output_dir, "params.json")
+        # save formatted json
+        with open(path, "w", encoding="utf-8") as f:
+            f.write(self.model_dump_json(indent=4))

competitions/runner.py

@@ -0,0 +1,88 @@
+import glob
+import json
+import os
+import time
+from dataclasses import dataclass
+
+import pandas as pd
+from huggingface_hub import snapshot_download
+from loguru import logger
+
+from competitions.info import CompetitionInfo
+from competitions.utils import run_evaluation
+
+
+@dataclass
+class JobRunner:
+    competition_info: CompetitionInfo
+    token: str
+    output_path: str
+
+    def __post_init__(self):
+        self.competition_id = self.competition_info.competition_id
+        self.competition_type = self.competition_info.competition_type
+        self.metric = self.competition_info.metric
+        self.submission_id_col = self.competition_info.submission_id_col
+        self.submission_cols = self.competition_info.submission_cols
+        self.submission_rows = self.competition_info.submission_rows
+
+    def get_pending_subs(self):
+        user_jsons = snapshot_download(
+            repo_id=self.competition_id,
+            allow_patterns="submission_info/*.json",
+            token=self.token,
+            repo_type="dataset",
+        )
+        user_jsons = glob.glob(os.path.join(user_jsons, "submission_info/*.json"))
+        pending_submissions = []
+        for _json in user_jsons:
+            _json = json.load(open(_json, "r", encoding="utf-8"))
+            user_id = _json["id"]
+            for sub in _json["submissions"]:
+                # if sub["status"] == "pending":
+                pending_submissions.append(
+                    {
+                        "user_id": user_id,
+                        "submission_id": sub["submission_id"],
+                        "date": sub["date"],
+                        "time": sub["time"],
+                    }
+                )
+        if len(pending_submissions) == 0:
+            logger.info("No pending submissions.")
+            return None
+        logger.info(f"Found {len(pending_submissions)} pending submissions.")
+        pending_submissions = pd.DataFrame(pending_submissions)
+        pending_submissions = pending_submissions.sort_values(by=["date", "time"])
+        pending_submissions = pending_submissions.reset_index(drop=True)
+        return pending_submissions
+
+    def run_local(self, pending_submissions):
+        for _, row in pending_submissions.iterrows():
+            user_id = row["user_id"]
+            submission_id = row["submission_id"]
+            eval_params = {
+                "competition_id": self.competition_id,
+                "competition_type": self.competition_type,
+                "metric": self.metric,
+                "token": self.token,
+                "user_id": user_id,
+                "submission_id": submission_id,
+                "submission_id_col": self.submission_id_col,
+                "submission_cols": self.submission_cols,
+                "submission_rows": self.submission_rows,
+                "output_path": self.output_path,
+            }
+            eval_params = json.dumps(eval_params)
+            eval_pid = run_evaluation(eval_params, local=True, wait=True)
+            logger.info(f"New evaluation process started with pid {eval_pid}.")
+
+    def run(self):
+        while True:
+            pending_submissions = self.get_pending_subs()
+            if pending_submissions is None:
+                time.sleep(5)
+                continue
+            if self.competition_type == "generic":
+                self.run_local(pending_submissions)
+            time.sleep(5)

competitions/submissions.py

@@ -22,7 +22,7 @@ class Submissions:
 
     def __post_init__(self):
         self.public_sub_columns = [
-            "date",
+            "datetime",
             "submission_id",
             "public_score",
             "submission_comment",
@@ -30,7 +30,7 @@ class Submissions:
             "status",
         ]
         self.private_sub_columns = [
-            "date",
+            "datetime",
             "submission_id",
             "public_score",
             "private_score",
@@ -326,16 +326,15 @@ class Submissions:
             raise SubmissionLimitError("Submission limit reached")
 
         logger.info(type(uploaded_file))
+        bytes_data = uploaded_file.file.read()
 
-        with open(uploaded_file.name, "rb") as f:
-            bytes_data = f.read()
         # verify file is valid
         if not self._verify_submission(bytes_data):
             raise SubmissionError("Invalid submission file")
         else:
             user_id = user_info["id"]
             submission_id = str(uuid.uuid4())
-            file_extension = uploaded_file.orig_name.split(".")[-1]
+            file_extension = uploaded_file.filename.split(".")[-1]
             # upload file to hf hub
             api = HfApi(token=self.token)
             api.upload_file(

competitions/templates/index.html

@@ -313,10 +313,11 @@
                 </div>
                 {% endif %}
                 <div class="form-group mt-2">
-                    <label for="params" class="text-sm font-medium text-gray-700">Submission description (optional)
+                    <label for="submission_comment" class="text-sm font-medium text-gray-700">Submission description
+                        (optional)
                     </label>
-                    <textarea id="params" name="params" rows="5" class="p-2.5 w-full text-sm text-gray-900"
-                        placeholder=""></textarea>
+                    <textarea id="submission_comment" name="submission_comment" rows="5"
+                        class="p-2.5 w-full text-sm text-gray-900" placeholder=""></textarea>
                 </div>
                 <div class="form-actions mt-6">
                     <button data-modal-hide="submission-modal" type="button"
@@ -356,10 +357,12 @@
                 return;
             }
 
-            // Token should be added here if available
             var token = document.getElementById('user_token').value;
             formData.append('token', token);
 
+            var submissionComment = document.getElementById('submission_comment').value;
+            formData.append('submission_comment', submissionComment);
+
             fetch('/new_submission', {
                 method: 'POST',
                 body: formData

competitions/utils.py

@@ -1,33 +1,16 @@
+import io
+import json
+import os
+import subprocess
+import traceback
+
 import requests
+from huggingface_hub import HfApi, hf_hub_download
 from loguru import logger
 
-from . import MOONLANDING_URL
-
-
-def get_auth_headers(token: str, prefix: str = "Bearer"):
-    return {"Authorization": f"{prefix} {token}"}
-
-
-def http_post(path: str, token: str, payload=None, domain: str = None, params=None) -> requests.Response:
-    """HTTP POST request to the AutoNLP API, raises UnreachableAPIError if the API cannot be reached"""
-    try:
-        response = requests.post(
-            url=domain + path, json=payload, headers=get_auth_headers(token=token), allow_redirects=True, params=params
-        )
-    except requests.exceptions.ConnectionError:
-        logger.error("❌ Failed to reach AutoNLP API, check your internet connection")
-    response.raise_for_status()
-    return response
+from competitions.params import EvalParams
 
-
-def http_get(path: str, token: str, domain: str = None) -> requests.Response:
-    """HTTP POST request to the AutoNLP API, raises UnreachableAPIError if the API cannot be reached"""
-    try:
-        response = requests.get(url=domain + path, headers=get_auth_headers(token=token), allow_redirects=True)
-    except requests.exceptions.ConnectionError:
-        logger.error("❌ Failed to reach AutoNLP API, check your internet connection")
-    response.raise_for_status()
-    return response
+from . import MOONLANDING_URL
 
 
 def user_authentication(token):
@@ -53,3 +36,99 @@ def user_authentication(token):
 def make_clickable_user(user_id):
     link = "https://huggingface.co/" + user_id
     return f'<a  target="_blank" href="{link}">{user_id}</a>'
+
+
+def run_evaluation(params, local=False, wait=False):
+    params = json.loads(params)
+    if isinstance(params, str):
+        params = json.loads(params)
+    params = EvalParams(**params)
+    if not local:
+        params.output_path = "/tmp/model"
+    params.save(output_dir=params.output_path)
+    cmd = [
+        "python",
+        "-m",
+        "competitions.evaluate",
+        "--config",
+        os.path.join(params.output_path, "params.json"),
+    ]
+
+    cmd = [str(c) for c in cmd]
+    logger.info(cmd)
+    env = os.environ.copy()
+    process = subprocess.Popen(" ".join(cmd), shell=True, env=env)
+    if wait:
+        process.wait()
+    return process.pid
+
+
+def pause_space(params):
+    if "SPACE_ID" in os.environ:
+        logger.info("Pausing space...")
+        api = HfApi(token=params.token)
+        api.pause_space(repo_id=os.environ["SPACE_ID"])
+
+
+def download_submission_info(params):
+    user_fname = hf_hub_download(
+        repo_id=params.competition_id,
+        filename=f"submission_info/{params.user_id}.json",
+        token=params.token,
+        repo_type="dataset",
+    )
+    with open(user_fname, "r", encoding="utf-8") as f:
+        user_submission_info = json.load(f)
+
+    return user_submission_info
+
+
+def upload_submission_info(params, user_submission_info):
+    user_submission_info_json = json.dumps(user_submission_info, indent=4)
+    user_submission_info_json_bytes = user_submission_info_json.encode("utf-8")
+    user_submission_info_json_buffer = io.BytesIO(user_submission_info_json_bytes)
+    api = HfApi(token=params.token)
+    api.upload_file(
+        path_or_fileobj=user_submission_info_json_buffer,
+        path_in_repo=f"submission_info/{params.user_id}.json",
+        repo_id=params.competition_id,
+        repo_type="dataset",
+    )
+
+
+def update_submission_status(params, status):
+    user_submission_info = download_submission_info(params)
+    for submission in user_submission_info["submissions"]:
+        if submission["submission_id"] == params.submission_id:
+            submission["status"] = status
+            break
+    upload_submission_info(params, user_submission_info)
+
+
+def update_submission_score(params, public_score, private_score):
+    user_submission_info = download_submission_info(params)
+    for submission in user_submission_info["submissions"]:
+        if submission["submission_id"] == params.submission_id:
+            submission["public_score"] = public_score
+            submission["private_score"] = private_score
+            submission["status"] = "done"
+            break
+    upload_submission_info(params, user_submission_info)
+
+
+def monitor(func):
+    def wrapper(*args, **kwargs):
+        params = kwargs.get("params", None)
+        if params is None and len(args) > 0:
+            params = args[0]
+
+        try:
+            return func(*args, **kwargs)
+        except Exception as e:
+            error_message = f"""{func.__name__} has failed due to an exception: {traceback.format_exc()}"""
+            logger.error(error_message)
+            logger.error(str(e))
+            update_submission_status(params, "failed")
+            pause_space(params)
+
+    return wrapper