refactor(hpc): ruff checks

scripts/analysis/explore_tensorboard.py

@@ -21,13 +21,14 @@ except ImportError:
     print("Error: Missing dependency. Please run: pip install tensorboard")
     sys.exit(1)
 
+
 def explore_run(log_dir):
     """
     Extracts and displays a summary of scalar metrics from a TensorBoard log directory.
     """
-    print(f"\n{'='*20} Exploring Run {'='*20}")
+    print(f"\n{'=' * 20} Exploring Run {'=' * 20}")
     print(f"Directory: {log_dir}")
-    print(f"{'='*55}\n")
+    print(f"{'=' * 55}\n")
 
     if not os.path.exists(log_dir):
         print(f"Error: Directory '{log_dir}' does not exist.")
@@ -35,17 +36,20 @@ def explore_run(log_dir):
 
     # Initialize EventAccumulator
     # size_guidance=0 loads all data points for each tag.
-    ea = event_accumulator.EventAccumulator(log_dir, size_guidance={
+    ea = event_accumulator.EventAccumulator(
-        event_accumulator.SCALARS: 0,
+        log_dir,
-        event_accumulator.TENSORS: 0,
+        size_guidance={
-    })
+            event_accumulator.SCALARS: 0,
-    
+            event_accumulator.TENSORS: 0,
+        },
+    )
+
     print("Loading event files (this may take a moment for large runs)...")
     ea.Reload()
 
     tags = ea.Tags()
-    scalar_tags = tags.get('scalars', [])
+    scalar_tags = tags.get("scalars", [])
-    
+
     if not scalar_tags:
         print("No scalar metrics found in this directory.")
         return None
@@ -60,43 +64,48 @@ def explore_run(log_dir):
         events = ea.Scalars(tag)
         if not events:
             continue
-        
+
         values = [e.value for e in events]
         last_event = events[-1]
         data[tag] = values
-        
+
-        summary.append({
+        summary.append(
-            "Metric": tag,
+            {
-            "Steps": len(events),
+                "Metric": tag,
-            "Last Value": f"{last_event.value:.4f}",
+                "Steps": len(events),
-            "Max": f"{max(values):.4f}",
+                "Last Value": f"{last_event.value:.4f}",
-            "Min": f"{min(values):.4f}"
+                "Max": f"{max(values):.4f}",
-        })
+                "Min": f"{min(values):.4f}",
+            }
+        )
 
     # Display summary table formatted manually
-    summary = sorted(summary, key=lambda x: x['Metric'])
+    summary = sorted(summary, key=lambda x: x["Metric"])
     print(f"{'Metric':<30} {'Steps':>10} {'Last':>12} {'Max':>12} {'Min':>12}")
     print("-" * 80)
     for row in summary:
-        print(f"{row['Metric']:<30} {row['Steps']:>10} {row['Last Value']:>12} {row['Max']:>12} {row['Min']:>12}")
+        print(
+            f"{row['Metric']:<30} {row['Steps']:>10} {row['Last Value']:>12} "
+            f"{row['Max']:>12} {row['Min']:>12}"
+        )
 
     # Calculate and display global metadata
-    if 'charts/SPS' in data:
+    if "charts/SPS" in data:
-        sps_events = ea.Scalars('charts/SPS')
+        sps_events = ea.Scalars("charts/SPS")
         if len(sps_events) > 1:
             total_duration_hours = (sps_events[-1].wall_time - sps_events[0].wall_time) / 3600
             print(f"\nTotal Recorded Duration: {total_duration_hours:.2f} hours")
-            
+
             # Estimate completion if total_timesteps is available in hyperparameters
             try:
-                hp_tags = [t for t in tags.get('tensors', []) if 'hyperparameters' in t]
+                hp_tags = [t for t in tags.get("tensors", []) if "hyperparameters" in t]
                 if hp_tags:
                     hp_event = ea.Tensors(hp_tags[0])[0]
-                    hp_text = hp_event.tensor_proto.string_val[0].decode('utf-8')
+                    hp_text = hp_event.tensor_proto.string_val[0].decode("utf-8")
-                    if 'total_timesteps' in hp_text:
+                    if "total_timesteps" in hp_text:
-                        for line in hp_text.split('\n'):
+                        for line in hp_text.split("\n"):
-                            if 'total_timesteps' in line:
+                            if "total_timesteps" in line:
-                                target = int(line.split('|')[2].strip())
+                                target = int(line.split("|")[2].strip())
                                 current = ea.Scalars(scalar_tags[0])[-1].step
                                 percent = (current / target) * 100
                                 print(f"Progress: {current:,} / {target:,} steps ({percent:.1f}%)")
@@ -105,31 +114,30 @@ def explore_run(log_dir):
 
     return data
 
+
 def main():
-    parser = argparse.ArgumentParser(description="Clean, reproducible TensorBoard exploration tool.")
+    parser = argparse.ArgumentParser(description="Reproducible TensorBoard exploration tool.")
     parser.add_argument("log_dir", help="Path to the TensorBoard run directory.")
-    parser.add_argument("--csv", help="Optional: Path to save all scalar data as a CSV.", default=None)
+    parser.add_argument("--csv", help="Optional: Path to export scalar data to CSV.", default=None)
-    
+
     args = parser.parse_args()
-    
+
     scalar_data = explore_run(args.log_dir)
-    
+
     if args.csv and scalar_data:
         # Reloading for wall_time and steps
         ea = event_accumulator.EventAccumulator(args.log_dir).Reload()
-        with open(args.csv, mode='w', newline='') as f:
+        with open(args.csv, mode="w", newline="") as f:
             writer = csv.DictWriter(f, fieldnames=["tag", "step", "value", "wall_time"])
             writer.writeheader()
             for tag in scalar_data.keys():
                 for e in ea.Scalars(tag):
-                    writer.writerow({
+                    writer.writerow(
-                        "tag": tag, 
+                        {"tag": tag, "step": e.step, "value": e.value, "wall_time": e.wall_time}
-                        "step": e.step, 
+                    )
-                        "value": e.value, 
+
-                        "wall_time": e.wall_time
-                    })
-        
         print(f"\nData exported to: {args.csv}")
 
+
 if __name__ == "__main__":
     main()

scripts/export_hpc_requirements.py

@@ -53,17 +53,17 @@ def main() -> None:
     dep_dict: dict[str, str] = {}
     for dep in data.get("project", {}).get("dependencies", []):
         dep_dict[normalise(pkg_name(dep))] = dep
-    
+
     # Add cuda extras (takes precedence for HPC)
     optional_deps = data.get("project", {}).get("optional-dependencies", {})
     for group in ["cuda"]:
         for dep in optional_deps.get(group, []):
             dep_dict[normalise(pkg_name(dep))] = dep
-    
+
     deps = list(dep_dict.values())
 
     final_deps: list[str] = []
-    print(f"Checking dependencies against HPC module list...", file=sys.stderr)
+    print("Checking dependencies against HPC module list...", file=sys.stderr)
     for dep in deps:
         name = normalise(pkg_name(dep))
         # Smart check: if the package name is a substring of any loaded module name

src/brittle_star_project/environment/env_config.py

@@ -55,10 +55,11 @@ def from_file(path: str) -> tuple[MorphologyConfig, ArenaConfig, EnvConfig]:
     with open(path, "r") as f:
         if path.endswith(".yaml") or path.endswith(".yml"):
             import yaml
+
             config_dict = yaml.safe_load(f)
         else:
             config_dict = json.load(f)
-            
+
         morphology = MorphologyConfig(**config_dict.get("morphology", {}))
         arena = ArenaConfig(**config_dict.get("arena", {}))
         env = EnvConfig(**config_dict.get("env", {}))

src/train.py

@@ -52,10 +52,12 @@ def train(args: PPOArgs):
     args.batch_size = args.num_envs * args.num_steps
     args.minibatch_size = args.batch_size // args.num_minibatches
     args.num_iterations = args.total_timesteps // args.batch_size
-    
+
     # Try to get git short hash
     try:
-        git_hash = subprocess.check_output(["git", "rev-parse", "--short", "HEAD"]).decode("ascii").strip()
+        git_hash = (
+            subprocess.check_output(["git", "rev-parse", "--short", "HEAD"]).decode("ascii").strip()
+        )
     except Exception:
         git_hash = "none"
 
@@ -64,8 +66,9 @@ def train(args: PPOArgs):
 
     if args.run_dir is None:
         args.run_dir = f"runs/{run_name}"
-    
+
     import os
+
     os.makedirs(args.run_dir, exist_ok=True)
 
     if args.track:
@@ -348,14 +351,14 @@ def train(args: PPOArgs):
             remaining_steps = args.total_timesteps - global_step
             eta_seconds = int(remaining_steps / sps) if sps > 0 else 0
             eta_str = str(datetime.timedelta(seconds=eta_seconds))
-            
+
             print(
                 f"Iteration {iteration}/{args.num_iterations} | "
                 f"Step {global_step}/{args.total_timesteps} | "
                 f"SPS {sps} | "
                 f"Return {avg_episodic_return:.4f} | "
-                f"ETA {eta_str}", 
+                f"ETA {eta_str}",
-                flush=True
+                flush=True,
             )
 
     if args.save_model:
@@ -388,7 +391,7 @@ def train(args: PPOArgs):
 
 def main() -> None:
     temp_args = tyro.cli(PPOArgs)
-    
+
     if temp_args.env_config_path is not None:
         with open(temp_args.env_config_path, "r") as f:
             config = yaml.safe_load(f)
@@ -397,7 +400,7 @@ def main() -> None:
                 for key, value in config.items():
                     if hasattr(temp_args, key):
                         setattr(temp_args, key, value)
-                
+
                 # Re-parse CLI to ensure they OVERRIDE the yaml
                 args = tyro.cli(PPOArgs, default=temp_args)
     else: