fix(lint): run ruff format on all files for CI compliance

sacredvoid · sacredvoid · commit ab8ee129945d · 2026-03-25T21:24:32.000-04:00
diff --git a/src/alignrl/eval.py b/src/alignrl/eval.py
@@ -46,8 +46,7 @@ def _resolve_preset(self) -> EvalConfig:
         if self.preset is not None:
             if self.preset not in BENCHMARK_PRESETS:
                 raise ValueError(
-                    f"Unknown preset {self.preset!r}. "
-                    f"Available: {', '.join(BENCHMARK_PRESETS)}"
+                    f"Unknown preset {self.preset!r}. Available: {', '.join(BENCHMARK_PRESETS)}"
                 )
             self.tasks = BENCHMARK_PRESETS[self.preset]
         else:
@@ -60,7 +59,8 @@ def parse_results(raw: dict[str, Any], model_name: str, stage: str) -> EvalResul
     benchmarks: dict[str, dict[str, float]] = {}
     for task_name, metrics in raw.get("results", {}).items():
         benchmarks[task_name] = {
-            k: v for k, v in metrics.items()
+            k: v
+            for k, v in metrics.items()
             if isinstance(v, (int, float)) and not isinstance(v, bool)
         }
     return EvalResult(model_name=model_name, stage=stage, benchmarks=benchmarks)
diff --git a/tests/test_cli.py b/tests/test_cli.py
@@ -45,12 +45,12 @@ def test_train_sft_stage(self, tmp_path) -> None:
         args = argparse.Namespace(config=str(config_path), stage="sft", push=None)
 
         mock_runner = MagicMock()
-        mock_runner.train.return_value = MagicMock(
-            output_dir=tmp_path, metrics={"train_loss": 0.5}
-        )
+        mock_runner.train.return_value = MagicMock(output_dir=tmp_path, metrics={"train_loss": 0.5})
 
-        with patch("alignrl.sft.SFTRunner", return_value=mock_runner) as mock_cls, \
-             patch("alignrl.sft.SFTConfig") as mock_cfg_cls:
+        with (
+            patch("alignrl.sft.SFTRunner", return_value=mock_runner) as mock_cls,
+            patch("alignrl.sft.SFTConfig") as mock_cfg_cls,
+        ):
             mock_cfg_cls.from_yaml.return_value = MagicMock()
             cmd_train(args)
             mock_cls.assert_called_once()
@@ -62,12 +62,12 @@ def test_train_grpo_stage(self, tmp_path) -> None:
         args = argparse.Namespace(config=str(config_path), stage="grpo", push=None)
 
         mock_runner = MagicMock()
-        mock_runner.train.return_value = MagicMock(
-            output_dir=tmp_path, metrics={"train_loss": 0.3}
-        )
+        mock_runner.train.return_value = MagicMock(output_dir=tmp_path, metrics={"train_loss": 0.3})
 
-        with patch("alignrl.grpo.GRPORunner", return_value=mock_runner), \
-             patch("alignrl.grpo.GRPOConfig") as mock_cfg_cls:
+        with (
+            patch("alignrl.grpo.GRPORunner", return_value=mock_runner),
+            patch("alignrl.grpo.GRPOConfig") as mock_cfg_cls,
+        ):
             mock_cfg_cls.from_yaml.return_value = MagicMock()
             cmd_train(args)
             mock_runner.train.assert_called_once()
@@ -78,12 +78,12 @@ def test_train_dpo_stage(self, tmp_path) -> None:
         args = argparse.Namespace(config=str(config_path), stage="dpo", push=None)
 
         mock_runner = MagicMock()
-        mock_runner.train.return_value = MagicMock(
-            output_dir=tmp_path, metrics={"train_loss": 0.2}
-        )
+        mock_runner.train.return_value = MagicMock(output_dir=tmp_path, metrics={"train_loss": 0.2})
 
-        with patch("alignrl.dpo.DPORunner", return_value=mock_runner), \
-             patch("alignrl.dpo.DPOConfig") as mock_cfg_cls:
+        with (
+            patch("alignrl.dpo.DPORunner", return_value=mock_runner),
+            patch("alignrl.dpo.DPOConfig") as mock_cfg_cls,
+        ):
             mock_cfg_cls.from_yaml.return_value = MagicMock()
             cmd_train(args)
             mock_runner.train.assert_called_once()
@@ -116,8 +116,10 @@ def test_eval_creates_output(self, tmp_path) -> None:
             wandb=False,
         )
 
-        with patch("alignrl.eval.EvalRunner", return_value=mock_runner), \
-             patch("alignrl.eval.EvalConfig"):
+        with (
+            patch("alignrl.eval.EvalRunner", return_value=mock_runner),
+            patch("alignrl.eval.EvalConfig"),
+        ):
             cmd_eval(args)
             mock_runner.evaluate.assert_called_once_with(stage="base")
             assert (tmp_path / "results" / "eval_base.json").exists()
diff --git a/tests/test_demo.py b/tests/test_demo.py
@@ -15,9 +15,11 @@ def test_creates_gradio_app(self) -> None:
         mock_server = MagicMock()
         mock_server.generate.return_value = "test output"
 
-        with patch.dict("sys.modules", {"gradio": mock_gr}), \
-             patch("alignrl.demo.ModelServer", return_value=mock_server), \
-             patch("alignrl.demo.InferenceConfig"):
+        with (
+            patch.dict("sys.modules", {"gradio": mock_gr}),
+            patch("alignrl.demo.ModelServer", return_value=mock_server),
+            patch("alignrl.demo.InferenceConfig"),
+        ):
             mock_server.load = MagicMock()
             app = create_demo(stages={"base": None}, model_name="test-model")
             assert app is not None
@@ -31,9 +33,11 @@ def test_multiple_stages(self) -> None:
 
         mock_server = MagicMock()
 
-        with patch.dict("sys.modules", {"gradio": mock_gr}), \
-             patch("alignrl.demo.ModelServer", return_value=mock_server), \
-             patch("alignrl.demo.InferenceConfig"):
+        with (
+            patch.dict("sys.modules", {"gradio": mock_gr}),
+            patch("alignrl.demo.ModelServer", return_value=mock_server),
+            patch("alignrl.demo.InferenceConfig"),
+        ):
             create_demo(
                 stages={"base": None, "sft": "./sft", "grpo": "./grpo"},
                 model_name="test-model",
@@ -65,9 +69,11 @@ def capture_click(fn, **kwargs):
         mock_button.click.side_effect = capture_click
         mock_gr.Button.return_value = mock_button
 
-        with patch.dict("sys.modules", {"gradio": mock_gr}), \
-             patch("alignrl.demo.ModelServer", return_value=mock_server), \
-             patch("alignrl.demo.InferenceConfig"):
+        with (
+            patch.dict("sys.modules", {"gradio": mock_gr}),
+            patch("alignrl.demo.ModelServer", return_value=mock_server),
+            patch("alignrl.demo.InferenceConfig"),
+        ):
             create_demo(stages={"base": None, "sft": "./sft"}, model_name="test")
 
         # Now call the captured respond_all function
diff --git a/tests/test_eval.py b/tests/test_eval.py
@@ -125,9 +125,7 @@ def test_save_results(self, tmp_path: Path) -> None:
         assert comparison["gsm8k"]["sft"]["exact_match"] == 0.45
 
     def test_save_results_creates_dir(self, tmp_path: Path) -> None:
-        result = EvalResult(
-            model_name="qwen", stage="base", benchmarks={"arc": {"acc": 0.5}}
-        )
+        result = EvalResult(model_name="qwen", stage="base", benchmarks={"arc": {"acc": 0.5}})
         cfg = EvalConfig()
         runner = EvalRunner(cfg)
         nested = tmp_path / "deep" / "nested" / "dir"
@@ -155,9 +153,7 @@ def test_evaluate_all_stages_restores_config(self) -> None:
         cfg = EvalConfig(adapter_path="original")
         runner = EvalRunner(cfg)
 
-        mock_result = EvalResult(
-            model_name="qwen", stage="base", benchmarks={}
-        )
+        mock_result = EvalResult(model_name="qwen", stage="base", benchmarks={})
 
         with patch.object(runner, "evaluate", return_value=mock_result):
             runner.evaluate_all_stages({"base": None, "sft": "./adapter"})
@@ -167,9 +163,7 @@ def test_evaluate_builds_model_args(self) -> None:
         cfg = EvalConfig(model_name="test-model", load_in_4bit=True, adapter_path="./adapter")
         runner = EvalRunner(cfg)
 
-        mock_raw = {
-            "results": {"gsm8k": {"exact_match,strict-match": 0.50}}
-        }
+        mock_raw = {"results": {"gsm8k": {"exact_match,strict-match": 0.50}}}
 
         mock_lm = MagicMock()
         mock_lm.simple_evaluate.return_value = mock_raw
