Modified template and got MVP up

Dockerfile

@@ -14,12 +14,21 @@ ENV HOME=/home/user \
 
 WORKDIR $HOME/app
 
+# Note(Lood): Mount with -v "$PWD:/home/user/app" to overwrite the COPY below
+
 RUN pip install --no-cache-dir --upgrade pip
-COPY --chown=user . $HOME/app
 
+# Enable quick rebuilds by only copying requirements in this layer
+COPY --chown=user requirements.txt $HOME/app
 RUN pip install -r requirements.txt
 
+COPY --chown=user . $HOME/app
+
 EXPOSE 7860
 ENV GRADIO_SERVER_NAME="0.0.0.0"
 
-CMD ["python", "app.py"]
+# CMD ["python", "app.py"]
+# Running with gradio enables hot reloading (https://www.gradio.app/guides/developing-faster-with-reload-mode)
+CMD ["gradio", "app.py"]
+# Or just run the container without the entrypoint and rerun the python app.py command when you want to reload
+

about.py

@@ -1,10 +1,10 @@
 import os
 from huggingface_hub import HfApi
 
-PROBLEM_TYPES = ["geometrical", "simple_to_build", "mhd_stable"]
+ASSAY_LIST = ["AC-SINS_pH7.4", "PSP_CHO", "HIC", "Tm2", "Titer"]
 TOKEN = os.environ.get("HF_TOKEN")
 CACHE_PATH=os.getenv("HF_HOME", ".")
 API = HfApi(token=TOKEN)
-organization="proxima-fusion"
-submissions_repo = f'{organization}/constellaration-bench-submissions'
-results_repo = f'{organization}/constellaration-bench-results'
+organization="ginkgo-datapoints"
+submissions_repo = f'{organization}/abdev-bench-submissions'
+results_repo = f'{organization}/abdev-bench-results'

app.py

@@ -1,22 +1,13 @@
-import pathlib
 from pathlib import Path
-import tempfile
-from typing import BinaryIO, Literal
 import json
 import pandas as pd
 
 import gradio as gr
-from datasets import load_dataset
-from gradio_leaderboard import ColumnFilter, Leaderboard, SelectColumns
+from gradio_leaderboard import Leaderboard
 from evaluation import evaluate_problem
-from datetime import datetime
-import os
 
-from submit import submit_boundary
-from about import PROBLEM_TYPES, TOKEN, CACHE_PATH, API, submissions_repo, results_repo
-from utils import read_submission_from_hub, read_result_from_hub, write_results, get_user, make_user_clickable, make_boundary_clickable
-from visualize import make_visual
-from evaluation import load_boundary, load_boundaries
+from utils import read_submission_from_hub, write_results
+from about import ASSAY_LIST
 
 def evaluate_boundary(filename):
     print(filename)
@@ -33,159 +24,152 @@ def evaluate_boundary(filename):
     write_results(data_dict, result)
     return
 
-def get_leaderboard():
-    ds = load_dataset(results_repo, split='train', download_mode="force_redownload")
-    full_df = pd.DataFrame(ds)
-    full_df['full results'] = full_df['result_filename'].apply(lambda x: make_boundary_clickable(x)).astype(str)
+def get_leaderboard_table(assay: str | None = None):
+    # ds = load_dataset(results_repo, split='train', download_mode="force_redownload")
+    # full_df = pd.DataFrame(ds)
+    # full_df['full results'] = full_df['result_filename'].apply(lambda x: make_boundary_clickable(x)).astype(str)
 
-    full_df.rename(columns={'submission_time': 'submission time', 'problem_type': 'problem type'}, inplace=True)
-    to_show = full_df.copy(deep=True)
-    to_show = to_show[to_show['user'] != 'test']
-    to_show = to_show[['submission time', 'problem type', 'user', 'score', 'full results']]
-    to_show['user'] = to_show['user'].apply(lambda x: make_user_clickable(x)).astype(str)
-
-    return to_show
+    # full_df.rename(columns={'submission_time': 'submission time', 'problem_type': 'problem type'}, inplace=True)
+    # to_show = full_df.copy(deep=True)
+    # to_show = to_show[to_show['user'] != 'test']
+    # to_show = to_show[['submission time', 'problem type', 'user', 'score', 'full results']]
+    # to_show['user'] = to_show['user'].apply(lambda x: make_user_clickable(x)).astype(str)
+    
+    # Previously hosted on HF hub, local for now
+    column_order = ["model", "feature", "assay", "spearman"]
+    df = pd.read_csv("data/metrics_all.csv")
+    if assay is not None:
+        df = df[df['assay'] == assay]
+    df = df[column_order]
+    print(df.head())
+    return df
+
+def get_leaderboard_object(assay: str | None = None):
+    df = get_leaderboard_table(assay=assay)
+    filter_columns = ["model"]
+    if assay is None:
+        filter_columns.append("assay")
+    Leaderboard(
+        value=df,
+        datatype=["str", "str", "str", "number"],
+        select_columns=["model", "assay", "feature", "spearman"],
+        # hide_columns=["spearman_abs"],
+        filter_columns=filter_columns,
+        every=60,
+        render=True
+    )
 
 def show_output_box(message):
     return gr.update(value=message, visible=True)
 
-def gradio_interface() -> gr.Blocks:
-    with gr.Blocks() as demo:
-        gr.Markdown("## Welcome to the ConStellaration Boundary Leaderboard!")
-        with gr.Tabs(elem_classes="tab-buttons"):
-            with gr.TabItem("🚀 Leaderboard", elem_id="boundary-benchmark-tab-table"):                
-                gr.Markdown("# Boundary Design Leaderboard")
-
-                Leaderboard(
-                    value=get_leaderboard(),
-                    datatype=['date', 'str', 'html', 'number', 'html'],
-                    select_columns=["submission time", "problem type", "user", "score", "full results"],
-                    search_columns=["submission time", "score", "user"],
-                    # hide_columns=["result_filename", "submission_filename", "objective", "minimize_objective", "boundary_json", "evaluated"],
-                    filter_columns=["problem type"],
-                    every=60,
-                    render=True
-                )
-
-                gr.Markdown("For the `geometrical` and `simple_to_build`, the scores are bounded between 0.0 and 1.0, where 1.0 is the best possible score. For the `mhd_stable` multi-objective problem, the score is unbounded with a undefined maximum score.")
-
-            with gr.TabItem("❔About", elem_id="boundary-benchmark-tab-table"):
-                gr.Markdown(
-                    """
-                    ## About This Challenge
-
-                    **Welcome to the ConStellaration Leaderboard**, a community-driven effort to accelerate fusion energy research using machine learning.
-
-                    In collaboration with [Proxima Fusion](https://www.proximafusion.com/), we're inviting the ML and physics communities to optimize plasma configurations for stellarators—a class of fusion reactors that offer steady-state operation and strong stability advantages over tokamaks.
-
-                    This leaderboard tracks submissions to a series of open benchmark tasks focused on:
-
-                    - **Geometrically optimized stellarators**
-                    - **Simple-to-build quasi-isodynamic (QI) stellarators**
-                    - **Multi-objective, MHD-stable QI stellarators**
-
-                    Participants are encouraged to build surrogate models, optimize plasma boundaries, and explore differentiable design pipelines that could replace or accelerate slow traditional solvers like VMEC++.
-
-                    ### Why It Matters
-
-                    Fusion promises clean, abundant, zero-carbon energy. But designing stellarators is computationally intense and geometrically complex. With open datasets, reference baselines, and your contributions, we can reimagine this process as fast, iterative, and ML-native.
-
-                    ### How to Participate
-
-                    - Clone the [ConStellaration dataset](https://huggingface.co/datasets/proxima-fusion/constellaration)
-                    - Build or train your model on the provided QI equilibria
-                    - Submit your predicted boundaries and results here to benchmark against others
-                    - Join the discussion and help expand the frontier of fusion optimization
-
-                    Let's bring fusion down to Earth—together.
-
-                """
-                )
-
-                # dropdown = gr.Dropdown(choices=filenames, label="Choose a file")
-                # plot_output = gr.Plot()
-
-            with gr.TabItem("🔍 Visualize", elem_id="boundary-benchmark-tab-table"):
-                ds = load_dataset(results_repo, split='train', download_mode="force_redownload")
-                full_df = pd.DataFrame(ds)
-                filenames = full_df['result_filename'].to_list()
-                with gr.Row():
-                    with gr.Column():
-                        dropdown = gr.Dropdown(choices=filenames, label="Choose a leaderboard entry", value=filenames[0])
-                        rld_btn = gr.Button(value="Reload")
-
-                    with gr.Column():
-                        plot = gr.Plot()
-
-                def get_boundary_vis(selected_file):
-                    local_path = read_result_from_hub(selected_file)
-                    with Path(local_path).open("r") as f:
-                        raw = f.read()
-                        data_dict = json.loads(raw)
-                        boundary_json = data_dict['boundary_json']
-
-                    if data_dict['problem_type'] == 'mhd_stable':
-                        raise gr.Error("Sorry this isn't implemented for mhd_stable submissions yet!")
-                    else:
-                        boundary = load_boundary(boundary_json)
-
-                    vis = make_visual(boundary)
-                    return vis
-
-                demo.load(get_boundary_vis, dropdown, plot)
-                rld_btn.click(get_boundary_vis, dropdown, plot)
-
-            with gr.TabItem("✉️ Submit", elem_id="boundary-benchmark-tab-table"):
-                gr.Markdown(
-                    """
-                # Plasma Boundary Evaluation Submission
-                Upload your plasma boundary JSON and select the problem type to get your score.
+# 
+# def gradio_interface() -> gr.Blocks:
+with gr.Blocks() as demo:
+    gr.Markdown("## Welcome to the Ginkgo Antibody Developability Benchmark Leaderboard!")
+    with gr.Tabs(elem_classes="tab-buttons"):
+        with gr.TabItem("🚀 Leaderboard", elem_id="abdev-benchmark-tab-table"):                
+            gr.Markdown("# Antibody Developability Benchmark Leaderboard")
+
+            get_leaderboard_object()
+
+            # gr.Markdown("Extra info here")
+
+        # Procedurally make these 5 tabs
+        for assay in ASSAY_LIST:
+            with gr.TabItem(assay, elem_id=f"abdev-benchmark-tab-table"):
+                gr.Markdown(f"# {assay}")
+                get_leaderboard_object(assay=assay)
+        
+        with gr.TabItem("❔About", elem_id="abdev-benchmark-tab-table"):
+            gr.Markdown(
                 """
-                )
-                filename = gr.State(value=None) 
-                eval_state = gr.State(value=None) 
-                user_state = gr.State(value=None)
-
-                # gr.LoginButton()
-
-                with gr.Row():
-                    with gr.Column():
-                        problem_type = gr.Dropdown(PROBLEM_TYPES, label="Problem Type")
-                        username_input = gr.Textbox(
-                            label="Username", 
-                            placeholder="Enter your Hugging Face username",
-                            info="This will be displayed on the leaderboard."
-                        )
-                    with gr.Column():
-                        boundary_file = gr.File(label="Boundary JSON File (.json)")
-
-                username_input.change(
-                    fn=lambda x: x if x.strip() else None,
-                    inputs=username_input,
-                    outputs=user_state
-                )                       
-
-                submit_btn = gr.Button("Evaluate")
-                message = gr.Textbox(label="Status", lines=1, visible=False)
-                # help message
-                gr.Markdown("If you have issues with submission or using the leaderboard, please start a discussion in the Community tab of this Space.")
-                
-                submit_btn.click(
-                    submit_boundary,
-                    inputs=[problem_type, boundary_file, user_state],
-                    outputs=[message, filename],
-                ).then(
-                    fn=show_output_box,
-                    inputs=[message],
-                    outputs=[message],
-                ).then(
-                    fn=evaluate_boundary,
-                    inputs=[filename],
-                    outputs=[eval_state]
-                )
- 
-    return demo
+                ## About
+                Some info here
+            """
+            )
+
+            # dropdown = gr.Dropdown(choices=filenames, label="Choose a file")
+            # plot_output = gr.Plot()
+
+        # with gr.TabItem("🔍 Visualize", elem_id="boundary-benchmark-tab-table"):
+        #     ds = load_dataset(results_repo, split='train', download_mode="force_redownload")
+        #     full_df = pd.DataFrame(ds)
+        #     filenames = full_df['result_filename'].to_list()
+        #     with gr.Row():
+        #         with gr.Column():
+        #             dropdown = gr.Dropdown(choices=filenames, label="Choose a leaderboard entry", value=filenames[0])
+        #             rld_btn = gr.Button(value="Reload")
+
+        #         with gr.Column():
+        #             plot = gr.Plot()
+
+        #     def get_boundary_vis(selected_file):
+        #         local_path = read_result_from_hub(selected_file)
+        #         with Path(local_path).open("r") as f:
+        #             raw = f.read()
+        #             data_dict = json.loads(raw)
+        #             boundary_json = data_dict['boundary_json']
+
+        #         if data_dict['problem_type'] == 'mhd_stable':
+        #             raise gr.Error("Sorry this isn't implemented for mhd_stable submissions yet!")
+        #         else:
+        #             boundary = load_boundary(boundary_json)
+
+        #         vis = make_visual(boundary)
+        #         return vis
+
+        #     demo.load(get_boundary_vis, dropdown, plot)
+        #     rld_btn.click(get_boundary_vis, dropdown, plot)
+
+        # with gr.TabItem("✉️ Submit", elem_id="boundary-benchmark-tab-table"):
+        #     gr.Markdown(
+        #         """
+        #     # Plasma Boundary Evaluation Submission
+        #     Upload your plasma boundary JSON and select the problem type to get your score.
+        #     """
+        #     )
+        #     filename = gr.State(value=None) 
+        #     eval_state = gr.State(value=None) 
+        #     user_state = gr.State(value=None)
+
+        #     # gr.LoginButton()
+
+        #     with gr.Row():
+        #         with gr.Column():
+        #             problem_type = gr.Dropdown(PROBLEM_TYPES, label="Problem Type")
+        #             username_input = gr.Textbox(
+        #                 label="Username", 
+        #                 placeholder="Enter your Hugging Face username",
+        #                 info="This will be displayed on the leaderboard."
+        #             )
+        #         with gr.Column():
+        #             boundary_file = gr.File(label="Boundary JSON File (.json)")
+
+        #     username_input.change(
+        #         fn=lambda x: x if x.strip() else None,
+        #         inputs=username_input,
+        #         outputs=user_state
+        #     )                       
+
+        #     submit_btn = gr.Button("Evaluate")
+        #     message = gr.Textbox(label="Status", lines=1, visible=False)
+        #     # help message
+        #     gr.Markdown("If you have issues with submission or using the leaderboard, please start a discussion in the Community tab of this Space.")
+            
+        #     submit_btn.click(
+        #         submit_boundary,
+        #         inputs=[problem_type, boundary_file, user_state],
+        #         outputs=[message, filename],
+        #     ).then(
+        #         fn=show_output_box,
+        #         inputs=[message],
+        #         outputs=[message],
+        #     ).then(
+        #         fn=evaluate_boundary,
+        #         inputs=[filename],
+        #         outputs=[eval_state]
+        #     )
 
 
 if __name__ == "__main__":
-    gradio_interface().launch()
+    demo.launch()

data/metrics_all.csv

@@ -0,0 +1,33 @@
+spearman,feature,assay,model,spearman_abs
+-0.358762795727933,PNC - tap,AC-SINS_pH7.4,tap,0.358762795727933
+-0.358762795727933,PNC - tap,AC-SINS_pH7.4,tap,0.358762795727933
+0.3203773185543964,SFvCSP - tap,AC-SINS_pH7.4,tap,0.3203773185543964
+0.3203773185543964,SFvCSP - tap,AC-SINS_pH7.4,tap,0.3203773185543964
+-0.2450651623577951,SFvCSP - tap,HIC,tap,0.2450651623577951
+0.2381972244142228,SFvCSP - tap,PR_CHO,tap,0.2381972244142228
+0.1923458958277369,CDR Length - tap,HIC,tap,0.1923458958277369
+-0.1839687842860175,PPC - tap,HIC,tap,0.1839687842860175
+0.1691412287169806,PPC - tap,AC-SINS_pH7.4,tap,0.1691412287169806
+0.1691412287169806,PPC - tap,AC-SINS_pH7.4,tap,0.1691412287169806
+-0.151234196032203,PNC - tap,PR_CHO,tap,0.151234196032203
+-0.1501689804134715,CDR Length - tap,AC-SINS_pH7.4,tap,0.1501689804134715
+-0.1501689804134715,CDR Length - tap,AC-SINS_pH7.4,tap,0.1501689804134715
+0.1423756688786398,PPC - tap,Titer,tap,0.1423756688786398
+0.1144051722170351,predictions,HIC,hic_mockup,0.1144051722170351
+0.1144051722170351,predictions,HIC,hic_mockup,0.1144051722170351
+-0.1043862069630446,PNC - tap,Titer,tap,0.1043862069630446
+0.0885325199884014,PPC - tap,PR_CHO,tap,0.0885325199884014
+0.0797623802296775,SFvCSP - tap,Titer,tap,0.0797623802296775
+-0.0729274519946903,CDR Length - tap,Tm2,tap,0.0729274519946903
+-0.0638862853735791,PSH - tap,Titer,tap,0.0638862853735791
+-0.0524556823713343,CDR Length - tap,PR_CHO,tap,0.0524556823713343
+-0.0482259337861972,PNC - tap,Tm2,tap,0.0482259337861972
+-0.0464732086283245,PSH - tap,HIC,tap,0.0464732086283245
+-0.0415639454664855,CDR Length - tap,Titer,tap,0.0415639454664855
+-0.0360881689253704,PSH - tap,AC-SINS_pH7.4,tap,0.0360881689253704
+-0.0360881689253704,PSH - tap,AC-SINS_pH7.4,tap,0.0360881689253704
+-0.0317350623491362,PSH - tap,Tm2,tap,0.0317350623491362
+0.0209381900169799,PSH - tap,PR_CHO,tap,0.0209381900169799
+-0.0179541460075494,SFvCSP - tap,Tm2,tap,0.0179541460075494
+0.0102627218889135,PNC - tap,HIC,tap,0.0102627218889135
+-0.0098872681569066,PPC - tap,Tm2,tap,0.0098872681569066

evaluation.py

@@ -1,42 +1,30 @@
 import json
 from pathlib import Path
 
-from constellaration import problems
-from constellaration.geometry import surface_rz_fourier
-
-PROBLEM_TYPES = ["geometrical", "simple_to_build", "mhd_stable"]
-
-def load_boundary(data: str) -> surface_rz_fourier.SurfaceRZFourier:
-    return surface_rz_fourier.SurfaceRZFourier.model_validate_json(data)
-
-def load_boundaries(data: str) -> list[surface_rz_fourier.SurfaceRZFourier]:
-    data_json = json.loads(data)
-    return [
-        surface_rz_fourier.SurfaceRZFourier.model_validate_json(b) for b in data_json
-    ]
-
 def evaluate_problem(
     problem_type: str, input_file: str
-) -> problems.EvaluationSingleObjective | problems.EvaluationMultiObjective:
-    with Path(input_file).open("r") as f:
-        raw = f.read()
-        data_dict = json.loads(raw)
-        data = data_dict['boundary_json']
+# ) -> problems.EvaluationSingleObjective | problems.EvaluationMultiObjective:
+):
+    pass
+    # with Path(input_file).open("r") as f:
+    #     raw = f.read()
+    #     data_dict = json.loads(raw)
+    #     data = data_dict['boundary_json']
 
-    print("Starting evaluation.")
+    # print("Starting evaluation.")
 
-    match problem_type:
-        case "geometrical":
-            boundary = load_boundary(data)
-            result = problems.GeometricalProblem().evaluate(boundary)
-        case "simple_to_build":
-            boundary = load_boundary(data)
-            result = problems.SimpleToBuildQIStellarator().evaluate(boundary)
-        case "mhd_stable":
-            boundaries = load_boundaries(data)
-            result = problems.MHDStableQIStellarator().evaluate(boundaries)
-        case _:
-            raise ValueError(f"Unknown problem type: {problem_type}")
+    # match problem_type:
+    #     case "geometrical":
+    #         boundary = load_boundary(data)
+    #         result = problems.GeometricalProblem().evaluate(boundary)
+    #     case "simple_to_build":
+    #         boundary = load_boundary(data)
+    #         result = problems.SimpleToBuildQIStellarator().evaluate(boundary)
+    #     case "mhd_stable":
+    #         boundaries = load_boundaries(data)
+    #         result = problems.MHDStableQIStellarator().evaluate(boundaries)
+    #     case _:
+    #         raise ValueError(f"Unknown problem type: {problem_type}")
         
-    print("Finished evaluation.")
-    return result
+    # print("Finished evaluation.")
+    # return result

requirements.txt

@@ -1,6 +1,5 @@
-constellaration==0.2.1
 gradio
 datasets
 huggingface_hub
 gradio-leaderboard
-plotly
+# plotly

utils.py

@@ -1,25 +1,19 @@
 import pathlib
-from pathlib import Path
 import tempfile
-from typing import BinaryIO, Literal
 import json
-import pandas as pd
 
 import gradio as gr
-from huggingface_hub import upload_file, hf_hub_download
-from evaluation import evaluate_problem
-from datetime import datetime
-import os
+from huggingface_hub import hf_hub_download
 
-from about import PROBLEM_TYPES, TOKEN, CACHE_PATH, API, submissions_repo, results_repo
+from about import API, submissions_repo, results_repo
 
-def make_user_clickable(name):
-    link =f'https://huggingface.co/{name}'
-    return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">{name}</a>'
+# def make_user_clickable(name):
+#     link =f'https://huggingface.co/{name}'
+#     return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">{name}</a>'
 
-def make_boundary_clickable(filename):
-    link =f'https://huggingface.co/datasets/proxima-fusion/constellaration-bench-results/blob/main/{filename}'
-    return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">link</a>'
+# def make_boundary_clickable(filename):
+#     link =f'https://huggingface.co/datasets/proxima-fusion/constellaration-bench-results/blob/main/{filename}'
+#     return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">link</a>'
 
 def read_result_from_hub(filename):
     local_path = hf_hub_download(

visualize.py

@@ -1,12 +0,0 @@
-from constellaration.utils import (
-    file_exporter,
-    visualization,
-    visualization_utils,
-)
-import gradio as gr
-import plotly.graph_objects as go
-
-def make_visual(boundary):
-    vis = visualization.plot_surface(boundary)
-    # div = vis.to_html(full_html=False)
-    return vis