Formatting & bug fixes

.github/workflows/pre-commit.yml

@@ -0,0 +1,26 @@
+name: pre-commit
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+jobs:
+  pre-commit:
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+
+      - name: Install pre-commit
+        run: |
+          python -m pip install --upgrade pip
+          pip install pre-commit
+
+      - name: Run pre-commit
+        run: pre-commit run --all-files

.gitignore

@@ -170,6 +170,7 @@ cython_debug/
 
 # Ruff stuff:
 .ruff_cache/
+.pre-commit-cache/
 
 # PyPI configuration file
 .pypirc

.pre-commit-config.yaml

@@ -0,0 +1,17 @@
+repos:
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.14.0
+    hooks:
+      - id: ruff
+        args:
+          - --fix
+          - --exit-non-zero-on-fix
+  - repo: local
+    hooks:
+      - id: ruff-format-with-kwargs
+        name: Ruff format with kwarg spacing
+        entry: scripts/run_ruff_format.py
+        language: python
+        types: [python]
+        additional_dependencies:
+          - ruff==0.6.9

pyproject.toml

@@ -48,7 +48,7 @@ huggingfacenotorch = [
     "tyro",
     "protobuf",
     "sentencepiece>=0.2.0",
-    "datasets>=3.4.1,!=4.0.*,!=4.1.0",
+    "datasets>=3.4.1,!=4.0.*,!=4.1.0,<4.4.0",
     "accelerate>=0.34.1",
     "peft>=0.7.1,!=0.11.0",
     "huggingface_hub>=0.34.0",
@@ -59,7 +59,7 @@ huggingfacenotorch = [
 ]
 huggingface = [
     "unsloth[huggingfacenotorch]",
-    "unsloth_zoo>=2025.11.1",
+    "unsloth_zoo>=2025.11.3",
     "torchvision",
     "unsloth[triton]",
 ]
@@ -489,11 +489,11 @@ colab-ampere-torch220 = [
     "flash-attn>=2.6.3 ; ('linux' in sys_platform)",
 ]
 colab-new = [
-    "unsloth_zoo>=2025.11.1",
+    "unsloth_zoo>=2025.11.3",
     "packaging",
     "tyro",
     "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,<=4.57.2",
-    "datasets>=3.4.1,!=4.0.*,!=4.1.0",
+    "datasets>=3.4.1,!=4.0.*,!=4.1.0,<4.4.0",
     "sentencepiece>=0.2.0",
     "tqdm",
     "psutil",
@@ -865,3 +865,32 @@ amd = [
 homepage = "http://www.unsloth.ai"
 documentation = "https://github.com/unslothai/unsloth"
 repository = "https://github.com/unslothai/unsloth"
+
+[tool.ruff]
+target-version = "py311"
+force-exclude = true
+extend-exclude = [
+    "*chat_templates.py",
+    "*ollama_template_mappers.py",
+    "*_auto_install.py",
+    "*mapper.py",
+]
+
+[tool.ruff.lint]
+select = ["E9", "F63", "F7", "F82"]
+ignore = [
+    "E402",
+    "E722",
+    "F403",
+    "F405",
+    "F811",
+    "F821",
+    "F841",
+    "F401",
+    "E731",
+    "E741",
+    "F601",
+    "E712",
+]
+
+[tool.ruff.format]

scripts/enforce_kwargs_spacing.py

@@ -0,0 +1,179 @@
+#!/usr/bin/env python3
+"""Ensure keyword arguments use spaces around '=', prune redundant pass statements."""
+
+from __future__ import annotations
+
+import ast
+import argparse
+import io
+import sys
+import tokenize
+from collections import defaultdict
+from pathlib import Path
+
+
+def enforce_spacing(text: str) -> tuple[str, bool]:
+    """Return updated text with keyword '=' padded by spaces, plus change flag."""
+    lines = text.splitlines(keepends=True)
+    if not lines:
+        return text, False
+
+    offsets: dict[int, int] = defaultdict(int)
+    changed = False
+
+    reader = io.StringIO(text).readline
+    for token in tokenize.generate_tokens(reader):
+        if token.type != tokenize.OP or token.string != "=":
+            continue
+
+        line_index = token.start[0] - 1
+        col = token.start[1] + offsets[line_index]
+
+        if line_index < 0 or line_index >= len(lines):
+            continue
+
+        line = lines[line_index]
+        if col >= len(line) or line[col] != "=":
+            continue
+
+        line_changed = False
+
+        # Insert a space before '=' when missing and not preceded by whitespace.
+        if col > 0 and line[col - 1] not in {" ", "\t"}:
+            line = f"{line[:col]} {line[col:]}"
+            offsets[line_index] += 1
+            col += 1
+            line_changed = True
+            changed = True
+
+        # Insert a space after '=' when missing and not followed by whitespace or newline.
+        next_index = col + 1
+        if next_index < len(line) and line[next_index] not in {" ", "\t", "\n", "\r"}:
+            line = f"{line[:next_index]} {line[next_index:]}"
+            offsets[line_index] += 1
+            line_changed = True
+            changed = True
+
+        if line_changed:
+            lines[line_index] = line
+
+    if not changed:
+        return text, False
+
+    return "".join(lines), True
+
+
+def remove_redundant_passes(text: str) -> tuple[str, bool]:
+    """Drop pass statements that share a block with other executable code."""
+
+    try:
+        tree = ast.parse(text)
+    except SyntaxError:
+        return text, False
+
+    redundant: list[ast.Pass] = []
+
+    def visit(node: ast.AST) -> None:
+        for attr in ("body", "orelse", "finalbody"):
+            value = getattr(node, attr, None)
+            if not isinstance(value, list) or len(value) <= 1:
+                continue
+            for stmt in value:
+                if isinstance(stmt, ast.Pass):
+                    redundant.append(stmt)
+            for stmt in value:
+                if isinstance(stmt, ast.AST):
+                    visit(stmt)
+        handlers = getattr(node, "handlers", None)
+        if handlers:
+            for handler in handlers:
+                visit(handler)
+
+    visit(tree)
+
+    if not redundant:
+        return text, False
+
+    lines = text.splitlines(keepends=True)
+    changed = False
+
+    for node in sorted(
+        redundant, key=lambda item: (item.lineno, item.col_offset), reverse=True
+    ):
+        start = node.lineno - 1
+        end = (node.end_lineno or node.lineno) - 1
+        if start >= len(lines):
+            continue
+        changed = True
+        if start == end:
+            line = lines[start]
+            col_start = node.col_offset
+            col_end = node.end_col_offset or (col_start + 4)
+            segment = line[:col_start] + line[col_end:]
+            lines[start] = segment if segment.strip() else ""
+            continue
+
+        # Defensive fall-back for unexpected multi-line 'pass'.
+        prefix = lines[start][: node.col_offset]
+        lines[start] = prefix if prefix.strip() else ""
+        for idx in range(start + 1, end):
+            lines[idx] = ""
+        suffix = lines[end][(node.end_col_offset or 0) :]
+        lines[end] = suffix
+
+    # Normalise to ensure lines end with newlines except at EOF.
+    result_lines: list[str] = []
+    for index, line in enumerate(lines):
+        if not line:
+            continue
+        if index < len(lines) - 1 and not line.endswith("\n"):
+            result_lines.append(f"{line}\n")
+        else:
+            result_lines.append(line)
+
+    return "".join(result_lines), changed
+
+
+def process_file(path: Path) -> bool:
+    try:
+        with tokenize.open(path) as handle:
+            original = handle.read()
+            encoding = handle.encoding
+    except (OSError, SyntaxError) as exc:  # SyntaxError from tokenize on invalid python
+        print(f"Failed to read {path}: {exc}", file=sys.stderr)
+        return False
+
+    updated, changed = enforce_spacing(original)
+    updated, removed = remove_redundant_passes(updated)
+    if changed or removed:
+        path.write_text(updated, encoding=encoding)
+        return True
+    return False
+
+
+def main(argv: list[str]) -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("files", nargs="+", help="Python files to fix")
+    args = parser.parse_args(argv)
+
+    touched: list[Path] = []
+    self_path = Path(__file__).resolve()
+
+    for entry in args.files:
+        path = Path(entry)
+        # Skip modifying this script to avoid self-edit loops.
+        if path.resolve() == self_path:
+            continue
+        if not path.exists() or path.is_dir():
+            continue
+        if process_file(path):
+            touched.append(path)
+
+    if touched:
+        for path in touched:
+            print(f"Adjusted kwarg spacing in {path}")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main(sys.argv[1:]))

scripts/run_ruff_format.py

@@ -0,0 +1,30 @@
+#!/usr/bin/env python3
+"""Run `ruff format` followed by kwarg spacing enforcement."""
+
+from __future__ import annotations
+
+import subprocess
+import sys
+from pathlib import Path
+
+HERE = Path(__file__).resolve().parent
+
+
+def main(argv: list[str]) -> int:
+    files = [arg for arg in argv if Path(arg).exists()]
+    if not files:
+        return 0
+
+    ruff_cmd = [sys.executable, "-m", "ruff", "format", *files]
+    ruff_proc = subprocess.run(ruff_cmd)
+    if ruff_proc.returncode != 0:
+        return ruff_proc.returncode
+
+    spacing_script = HERE / "enforce_kwargs_spacing.py"
+    spacing_cmd = [sys.executable, str(spacing_script), *files]
+    spacing_proc = subprocess.run(spacing_cmd)
+    return spacing_proc.returncode
+
+
+if __name__ == "__main__":
+    raise SystemExit(main(sys.argv[1:]))