chore: ruff lint cleanup (auto-fixable + manual)

Fixes all ruff errors flagged by CI:
- I001: import ordering (auto-fixed)
- F401: unused imports removed
- F841: unused pipeline variable removed from cmd_publish
- B905: explicit strict=False on zips where lengths intentionally vary
- B017: narrower AttributeError instead of bare Exception for frozen dataclass test
- E501: wrapped long NotImplementedError signatures in trainer.py

37 tests still pass. Ruff check src tests clean.

Author: Ronald Tse

src/cli.py

@@ -132,14 +132,6 @@ def cmd_export(args: argparse.Namespace) -> int:
 
 
 def cmd_publish(args: argparse.Namespace) -> int:
-    from framework.pipeline import TrainingPipeline
-
-    pipeline = TrainingPipeline.from_config(
-        task_name=args.task,
-        data_root=args.data_root,
-        out_root=args.out_root,
-        tasks_root=args.tasks_root,
-    )
     repo_id = args.repo or f"interscript/{args.task}"
     print(f"Would upload {args.out_root} to huggingface.co/{repo_id}")
     return 0

src/framework/__init__.py

@@ -8,19 +8,19 @@
 
 from framework.config import TaskConfig, load_task_config
 from framework.data import DataModule, DataSplit, Example
-from framework.model import ModelModule, ModelKind
-from framework.trainer import BaseTrainer, TrainingState
 from framework.evaluator import BaseEvaluator, MetricSet
-from framework.exporter import OnnxExporter, ExportResult
+from framework.exporter import ExportResult, OnnxExporter
+from framework.model import ModelKind, ModelModule
 from framework.pipeline import TrainingPipeline
 from framework.registry import (
     register_data_module,
-    register_model_module,
     register_evaluator,
+    register_model_module,
     resolve_data_module,
-    resolve_model_module,
     resolve_evaluator,
+    resolve_model_module,
 )
+from framework.trainer import BaseTrainer, TrainingState
 
 __all__ = [
     "TaskConfig",

src/framework/config.py

@@ -26,7 +26,7 @@ class DataConfig:
     cleaner: str = "basic"
 
     @classmethod
-    def from_dict(cls, raw: dict[str, Any]) -> "DataConfig":
+    def from_dict(cls, raw: dict[str, Any]) -> DataConfig:
         return cls(
             module=raw["module"],
             source=raw["source"],
@@ -50,7 +50,7 @@ class ModelConfig:
     lora_alpha: int = 32
 
     @classmethod
-    def from_dict(cls, raw: dict[str, Any]) -> "ModelConfig":
+    def from_dict(cls, raw: dict[str, Any]) -> ModelConfig:
         return cls(
             module=raw["module"],
             teacher_name=raw["teacher_name"],
@@ -80,7 +80,7 @@ class TrainConfig:
     out_dir: str = "models"
 
     @classmethod
-    def from_dict(cls, raw: dict[str, Any]) -> "TrainConfig":
+    def from_dict(cls, raw: dict[str, Any]) -> TrainConfig:
         return cls(
             epochs=raw.get("epochs", 3),
             batch_size=raw.get("batch_size", 16),
@@ -106,7 +106,7 @@ class EvalConfig:
     batch_size: int = 32
 
     @classmethod
-    def from_dict(cls, raw: dict[str, Any]) -> "EvalConfig":
+    def from_dict(cls, raw: dict[str, Any]) -> EvalConfig:
         return cls(
             module=raw["module"],
             metric=raw["metric"],
@@ -126,7 +126,7 @@ class ExportConfig:
     quantize: bool = False
 
     @classmethod
-    def from_dict(cls, raw: dict[str, Any]) -> "ExportConfig":
+    def from_dict(cls, raw: dict[str, Any]) -> ExportConfig:
         return cls(
             opset=raw.get("opset", 17),
             dynamic_axes=raw.get("dynamic_axes", {"input_ids": {0: "batch", 1: "seq"}}),
@@ -152,7 +152,7 @@ class TaskConfig:
     export: ExportConfig
 
     @classmethod
-    def from_dict(cls, name: str, raw: dict[str, Any]) -> "TaskConfig":
+    def from_dict(cls, name: str, raw: dict[str, Any]) -> TaskConfig:
         return cls(
             name=name,
             description=raw.get("description", ""),

src/framework/data.py

@@ -12,9 +12,9 @@
 from __future__ import annotations
 
 from abc import ABC, abstractmethod
+from collections.abc import Iterator, Sequence
 from dataclasses import dataclass
 from pathlib import Path
-from typing import Iterator, Sequence
 
 from framework.config import DataConfig
 

src/framework/evaluator.py

@@ -12,8 +12,8 @@
 
 from abc import ABC, abstractmethod
 from collections import Counter
+from collections.abc import Sequence
 from dataclasses import dataclass, field
-from typing import Sequence
 
 
 @dataclass(frozen=True)
@@ -110,7 +110,7 @@ def accuracy(pred: Sequence[str], gold: Sequence[str]) -> float:
     """Exact-match accuracy across the corpus."""
     if not pred:
         return 0.0
-    correct = sum(1 for p, g in zip(pred, gold) if p == g)
+    correct = sum(1 for p, g in zip(pred, gold, strict=False) if p == g)
     return correct / len(pred)
 
 
@@ -121,8 +121,8 @@ def most_common_error_pairs(
 ) -> dict[str, float]:
     """Frequency of (gold, pred) character-level substitutions."""
     counts: Counter[tuple[str, str]] = Counter()
-    for p, g in zip(pred, gold):
-        for pc, gc in zip(p, g):
+    for p, g in zip(pred, gold, strict=False):
+        for pc, gc in zip(p, g, strict=False):
             if pc != gc:
                 counts[(gc, pc)] += 1
     total = sum(counts.values())

src/framework/model.py

@@ -12,9 +12,10 @@
 from __future__ import annotations
 
 from abc import ABC, abstractmethod
+from collections.abc import Sequence
 from dataclasses import dataclass
 from enum import Enum
-from typing import Any, Sequence
+from typing import Any
 
 
 class ModelKind(str, Enum):

src/framework/pipeline.py

@@ -9,22 +9,21 @@
 from __future__ import annotations
 
 import importlib
+from collections.abc import Callable
 from dataclasses import dataclass
 from pathlib import Path
-from typing import Any, Callable
 
 from framework.config import TaskConfig
 from framework.data import DataModule
 from framework.evaluator import BaseEvaluator, MetricSet
 from framework.exporter import ExportResult, OnnxExporter
 from framework.model import ModelModule
-from framework.trainer import BaseTrainer, TrainConfig
 from framework.registry import (
     resolve_data_module,
     resolve_evaluator,
     resolve_model_module,
 )
-
+from framework.trainer import BaseTrainer
 
 TrainerFactory = Callable[..., BaseTrainer]
 
@@ -73,7 +72,7 @@ def __init__(
         model_class: type[ModelModule] | None = None,
         evaluator_class: type[BaseEvaluator] | None = None,
         exporter: OnnxExporter | None = None,
-        trainer_factory: "TrainerFactory | None" = None,
+        trainer_factory: TrainerFactory | None = None,
     ) -> None:
         self.config = config
         self.data_root = data_root
@@ -92,7 +91,7 @@ def from_config(
         data_root: Path,
         out_root: Path,
         tasks_root: Path | None = None,
-    ) -> "TrainingPipeline":
+    ) -> TrainingPipeline:
         """Build a pipeline from ``src/tasks/<task_name>/config.yaml``."""
         from framework.config import load_task_config
 
@@ -139,7 +138,6 @@ def run(
 
         model = self.build_model()
 
-        from framework.trainer import BaseTrainer
 
         trainer = self._construct_trainer(model, data)
         state = trainer.fit(max_steps=max_steps)

src/framework/trainer.py

@@ -18,7 +18,7 @@
 from abc import ABC, abstractmethod
 from dataclasses import dataclass, field
 from pathlib import Path
-from typing import Any, Callable
+from typing import Any
 
 from framework.config import TrainConfig
 from framework.data import DataModule
@@ -148,11 +148,15 @@ def _checkpoint(self, tag: str) -> None:
 class FineTuneTrainer(BaseTrainer):
     """Teacher trainer: supervised CE loss on gold labels."""
 
-    def compute_loss(self, batch: Any) -> tuple[Any, dict[str, float]]:  # pragma: no cover - torch-bound
-        raise NotImplementedError("FineTuneTrainer requires torch; implement in tasks that use it")
+    def compute_loss(self, batch):  # pragma: no cover - torch-bound
+        raise NotImplementedError(
+            "FineTuneTrainer requires torch; implement in tasks that use it"
+        )
 
     def make_optimizer(self) -> Any:  # pragma: no cover - torch-bound
-        raise NotImplementedError("FineTuneTrainer requires torch; implement in tasks that use it")
+        raise NotImplementedError(
+            "FineTuneTrainer requires torch; implement in tasks that use it"
+        )
 
 
 class DistillTrainer(BaseTrainer):
@@ -174,8 +178,10 @@ def __init__(
         super().__init__(config, model, data, out_dir)
         self.teacher = teacher
 
-    def compute_loss(self, batch: Any) -> tuple[Any, dict[str, float]]:  # pragma: no cover - torch-bound
-        raise NotImplementedError("DistillTrainer requires torch; implement in tasks that use it")
+    def compute_loss(self, batch):  # pragma: no cover - torch-bound
+        raise NotImplementedError(
+            "DistillTrainer requires torch; implement in tasks that use it"
+        )
 
     def make_optimizer(self) -> Any:  # pragma: no cover - torch-bound
         raise NotImplementedError("DistillTrainer requires torch; implement in tasks that use it")

src/tasks/rababa_arabic/data.py

@@ -12,11 +12,9 @@
 from __future__ import annotations
 
 import random
-from collections import Counter
+from collections.abc import Sequence
 from pathlib import Path
-from typing import Sequence
 
-from framework.config import DataConfig
 from framework.data import DataModule, DataSplit, Example, PreparedData
 from framework.registry import register_data_module
 

src/tasks/rababa_arabic/metrics.py

@@ -11,7 +11,7 @@
 
 from __future__ import annotations
 
-from typing import Sequence
+from collections.abc import Sequence
 
 from framework.evaluator import (
     BaseEvaluator,
@@ -30,7 +30,7 @@ class DEREvaluator(BaseEvaluator):
 
     def compute_metric(self, predictions: Sequence[str], gold: Sequence[str]) -> float:
         total = 0.0
-        for pred, ref in zip(predictions, gold):
+        for pred, ref in zip(predictions, gold, strict=False):
             total += char_error_rate(pred, ref)
         return total / len(predictions)