Update inference_app.py

inference_app.py

@@ -1,7 +1,6 @@
 from __future__ import annotations
 from pathlib import Path
 import time
-from biotite.application.autodock import VinaApp
 
 import gradio as gr
 
@@ -24,6 +23,31 @@ EVAL_METRICS = ["system", "LDDT-PLI", "LDDT-LP", "BISY-RMSD"]
 EVAL_METRICS_PINDER = ["system","L_rms", "I_rms", "F_nat", "DOCKQ", "CAPRI_class"]
 
 
+import os
+
+from huggingface_hub import HfApi
+
+# Info to change for your repository
+# ----------------------------------
+TOKEN = os.environ.get("HF_TOKEN") # A read/write token for your org
+
+OWNER = "MLSB" # Change to your org - don't forget to create a results and request dataset, with the correct format!
+# ----------------------------------
+
+REPO_ID = f"{OWNER}/leaderboard2024"
+QUEUE_REPO = f"{OWNER}/requests"
+RESULTS_REPO = f"{OWNER}/results"
+
+# If you setup a cache later, just change HF_HOME
+CACHE_PATH=os.getenv("HF_HOME", ".")
+
+# Local caches
+EVAL_REQUESTS_PATH = os.path.join(CACHE_PATH, "eval-queue")
+EVAL_RESULTS_PATH = os.path.join(CACHE_PATH, "eval-results")
+EVAL_REQUESTS_PATH_BACKEND = os.path.join(CACHE_PATH, "eval-queue-bk")
+EVAL_RESULTS_PATH_BACKEND = os.path.join(CACHE_PATH, "eval-results-bk")
+
+API = HfApi(token=TOKEN)
 
 
 
@@ -85,13 +109,55 @@ def get_metrics(
     return gr.DataFrame(metrics, visible=True), run_time
 
 
+
 def get_metrics_pinder(
     system_id: str,
     complex_file: Path,
     methodname: str = "",
     store:bool =True
 ) -> tuple[pd.DataFrame, float]:
-    return pd.DataFrame(), 0
+    start_time = time.time()
+    
+    if not isinstance(prediction, Path):
+        prediction = Path(prediction)
+    # Infer the ground-truth name from prediction filename or directory where its stored
+    # We need to figure out how we plan to consistently map predictions to systems so that eval metrics can be calculated
+    # I assume we won't distribute the ground-truth structures (though they are already accessible if we don't blind system IDs)
+    native = Path(f"./ground_truth/{system_id}.pdb")
+    # alternatively 
+    # native = Path(f"./ground_truth/{prediction.parent.parent.stem}.pdb")
+    # OR we need the user to provide prediction + system name
+    try:        
+        # Get eval metrics for the prediction
+        bdq = BiotiteDockQ(native, complex_file.name, parallel_io=False)
+        metrics = bdq.calculate()
+        metrics = metrics[["system", "LRMS", "iRMS", "Fnat", "DockQ", "CAPRI"]].copy()
+        metrics.rename(columns={"LRMS": "L_rms", "iRMS": "I_rms", "Fnat": "F_nat", "DockQ": "DOCKQ", "CAPRI": "CAPRI_class"}, inplace=True)
+    except Exception as e:
+        failed_metrics = {"L_rms": 100.0, "I_rms": 100.0, "F_nat": 0.0, "DOCKQ": 0.0, "CAPRI_class": "Incorrect"} 
+        metrics = pd.DataFrame([failed_metrics])
+        metrics["system"] = native.stem
+        gr.Error(f"Failed to evaluate prediction [{prediction}]:\n{e}")
+    # Upload to hub
+    with tempfile.NamedTemporaryFile as temp:
+        metrics.to_csv(temp.name)
+        API.upload_file(
+                path_or_fileobj=temp.name,
+                path_in_repo=f"{dataset}/{methodname}/{system_id}/",
+                repo_id=QUEUE_REPO,
+                repo_type="dataset",
+                commit_message=f"Add {model_name} to eval queue",
+        )
+    API.upload_file(
+            path_or_fileobj=complex_file.name,
+            path_in_repo=f"{dataset}/{methodname}/{system_id}/",
+            repo_id=QUEUE_REPO,
+            repo_type="dataset",
+            commit_message=f"Add {model_name} to eval queue",
+    )
+    end_time = time.time()
+    run_time = end_time - start_time
+    return gr.DataFrame(metrics, visible=True), run_time
 
 with gr.Blocks() as app:
     with gr.Tab("🧬 PINDER evaluation template"):
@@ -127,7 +193,7 @@ with gr.Blocks() as app:
                 posebusters = gr.Checkbox(label="PoseBusters", value=True)
                 methodname = gr.Textbox(label="Name of your method in the format mlsb/spacename")
                 store = gr.Checkbox(label="Store on huggingface for leaderboard", value=False)
-         gr.Examples(
+        gr.Examples(
             [
                 [
                     "4neh__1__1.B__1.H",