Use per-build variant content hashes in the lockfile

This makes the lock file a fair bit shorter than per-file hashes. They
can also be verified without any knowledge of Git objects, etc.

src/hf_kernels/cache.py

@@ -0,0 +1,4 @@
+from typing import Optional
+import os
+
+CACHE_DIR: Optional[str] = os.environ.get("HF_KERNELS_CACHE", None)

src/hf_kernels/cli.py

@@ -6,7 +6,7 @@
 
 from hf_kernels.compat import tomllib
 from hf_kernels.lockfile import KernelLock, get_kernel_locks
-from hf_kernels.utils import install_kernel, install_kernel_all_variants
+from hf_kernels.utils import build_variant, install_kernel, install_kernel_all_variants
 
 
 def main():
@@ -53,16 +53,22 @@ def download_kernels(args):
     all_successful = True
 
     for kernel_lock_json in lock_json:
-        kernel_lock = KernelLock.from_json(kernel_lock_json)
+        kernel_lock = KernelLock(**kernel_lock_json)
         print(
             f"Downloading `{kernel_lock.repo_id}` at with SHA: {kernel_lock.sha}",
             file=sys.stderr,
         )
         if args.all_variants:
-            install_kernel_all_variants(kernel_lock.repo_id, kernel_lock.sha)
+            install_kernel_all_variants(
+                kernel_lock.repo_id, kernel_lock.sha, hashes=kernel_lock.variants
+            )
         else:
             try:
-                install_kernel(kernel_lock.repo_id, kernel_lock.sha)
+                install_kernel(
+                    kernel_lock.repo_id,
+                    kernel_lock.sha,
+                    hash=kernel_lock.variants[build_variant()],
+                )
             except FileNotFoundError as e:
                 print(e, file=sys.stderr)
                 all_successful = False

src/hf_kernels/hash.py

@@ -0,0 +1,31 @@
+from typing import List, Tuple
+import hashlib
+import os
+from pathlib import Path
+
+
+def content_hash(dir: Path) -> str:
+    """Get a hash of the contents of a directory."""
+
+    # Get the file paths. The first element is a byte-encoded relative path
+    # used for sorting. The second element is the absolute path.
+    paths: List[Tuple[bytes, Path]] = []
+    # Ideally we'd use Path.walk, but it's only available in Python 3.12.
+    for dirpath, _, filenames in os.walk(dir):
+        for filename in filenames:
+            file_abs = Path(dirpath) / filename
+
+            # Python likes to create files when importing modules from the
+            # cache, only hash files that are symlinked blobs.
+            if file_abs.is_symlink():
+                paths.append(
+                    (file_abs.relative_to(dir).as_posix().encode("utf-8"), file_abs)
+                )
+
+    m = hashlib.sha256()
+    for filename, path in paths:
+        m.update(filename)
+        with open(path, "rb") as f:
+            m.update(f.read())
+
+    return f"sha256-{m.hexdigest()}"

src/hf_kernels/lockfile.py

@@ -1,30 +1,21 @@
 from dataclasses import dataclass
 from pathlib import Path
-from typing import Dict, List
+from typing import Dict
 
-from huggingface_hub import HfApi
+from huggingface_hub import HfApi, snapshot_download
 from packaging.specifiers import SpecifierSet
 from packaging.version import InvalidVersion, Version
 
 from hf_kernels.compat import tomllib
-
-
-@dataclass
-class FileLock:
-    filename: str
-    blob_id: str
+from hf_kernels.cache import CACHE_DIR
+from hf_kernels.hash import content_hash
 
 
 @dataclass
 class KernelLock:
     repo_id: str
     sha: str
-    files: List[FileLock]
-
-    @classmethod
-    def from_json(cls, o: Dict):
-        files = [FileLock(**f) for f in o["files"]]
-        return cls(repo_id=o["repo_id"], sha=o["sha"], files=files)
+    variants: Dict[str, str]
 
 
 def _get_available_versions(repo_id: str):
@@ -59,30 +50,23 @@ def get_kernel_locks(repo_id: str, version_spec: str):
 
     tag_for_newest = versions[accepted_versions[-1]]
 
-    r = HfApi().repo_info(
-        repo_id=repo_id, revision=tag_for_newest.target_commit, files_metadata=True
-    )
-    if r.sha is None:
-        raise ValueError(
-            f"Cannot get commit SHA for repo {repo_id} for tag {tag_for_newest.name}"
-        )
-
-    if r.siblings is None:
-        raise ValueError(
-            f"Cannot get sibling information for {repo_id} for tag {tag_for_newest.name}"
+    repo_path = Path(
+        snapshot_download(
+            repo_id,
+            allow_patterns="build/*",
+            cache_dir=CACHE_DIR,
+            revision=tag_for_newest.target_commit,
         )
+    )
 
-    file_locks = []
-    for sibling in r.siblings:
-        if sibling.rfilename.startswith("build/torch"):
-            if sibling.blob_id is None:
-                raise ValueError(f"Cannot get blob ID for {sibling.rfilename}")
-
-            file_locks.append(
-                FileLock(filename=sibling.rfilename, blob_id=sibling.blob_id)
-            )
+    variant_hashes = {}
+    for entry in (repo_path / "build").iterdir():
+        variant = entry.parts[-1]
+        variant_hashes[variant] = content_hash(entry)
 
-    return KernelLock(repo_id=repo_id, sha=r.sha, files=file_locks)
+    return KernelLock(
+        repo_id=repo_id, sha=tag_for_newest.target_commit, variants=variant_hashes
+    )
 
 
 def write_egg_lockfile(cmd, basename, filename):

src/hf_kernels/utils.py

@@ -14,11 +14,11 @@
 from huggingface_hub import hf_hub_download, snapshot_download
 from packaging.version import parse
 
+from hf_kernels.cache import CACHE_DIR
 from hf_kernels.compat import tomllib
+from hf_kernels.hash import content_hash
 from hf_kernels.lockfile import KernelLock
 
-CACHE_DIR: Optional[str] = os.environ.get("HF_KERNELS_CACHE", None)
-
 
 def build_variant():
     import torch
@@ -47,7 +47,10 @@ def import_from_path(module_name: str, file_path):
 
 
 def install_kernel(
-    repo_id: str, revision: str, local_files_only: bool = False
+    repo_id: str,
+    revision: str,
+    local_files_only: bool = False,
+    hash: Optional[str] = None,
 ) -> Tuple[str, str]:
     """Download a kernel for the current environment to the cache."""
     package_name = get_metadata(repo_id, revision, local_files_only=local_files_only)[
@@ -62,6 +65,14 @@ def install_kernel(
     )
 
     variant_path = f"{repo_path}/build/{build_variant()}"
+
+    if hash is not None:
+        found_hash = content_hash(Path(variant_path))
+        if found_hash != hash:
+            raise ValueError(
+                f"Expected hash {hash} for path {variant_path}, but got: {found_hash}"
+            )
+
     module_init_path = f"{variant_path}/{package_name}/__init__.py"
 
     if not os.path.exists(module_init_path):
@@ -73,16 +84,37 @@ def install_kernel(
 
 
 def install_kernel_all_variants(
-    repo_id: str, revision: str, local_files_only: bool = False
-):
-    snapshot_download(
-        repo_id,
-        allow_patterns="build/*",
-        cache_dir=CACHE_DIR,
-        revision=revision,
-        local_files_only=local_files_only,
+    repo_id: str,
+    revision: str,
+    local_files_only: bool = False,
+    hashes: Optional[dict] = None,
+) -> str:
+    repo_path = Path(
+        snapshot_download(
+            repo_id,
+            allow_patterns="build/*",
+            cache_dir=CACHE_DIR,
+            revision=revision,
+            local_files_only=local_files_only,
+        )
     )
 
+    for entry in (repo_path / "build").iterdir():
+        variant = entry.parts[-1]
+
+        if hashes is not None:
+            hash = hashes.get(variant)
+            if hash is None:
+                raise ValueError(f"No hash found for build variant: {variant}")
+
+            found_hash = content_hash(entry)
+            if found_hash != hash:
+                raise ValueError(
+                    f"Expected hash {hash} for path {entry}, but got: {found_hash}"
+                )
+
+    return f"{repo_path}/build"
+
 
 def get_metadata(repo_id: str, revision: str, local_files_only: bool = False):
     with open(
@@ -145,7 +177,7 @@ def _get_caller_locked_kernel(repo_id: str) -> Optional[str]:
         lock_json = dist.read_text("hf-kernels.lock")
         if lock_json is not None:
             for kernel_lock_json in json.loads(lock_json):
-                kernel_lock = KernelLock.from_json(kernel_lock_json)
+                kernel_lock = KernelLock(**kernel_lock_json)
                 if kernel_lock.repo_id == repo_id:
                     return kernel_lock.sha
     return None

tests/hash_validation/hf-kernels.lock

@@ -0,0 +1,20 @@
+[
+  {
+    "repo_id": "kernels-community/activation",
+    "sha": "6a030420d0dd33ffdc1281afc8ae8e94b4f4f9d0",
+    "variants": {
+      "torch25-cxx98-cu124-x86_64-linux": "sha256-1f1acbf45bb98cc71d0d082e19c40d6a7be57a5ccebf1143539c91ee38a5ec1c",
+      "torch25-cxx11-cu121-x86_64-linux": "sha256-0cf715a909a5218e62ed33dd941a46038ebf5b444163d70eb25e2f611f67bf51",
+      "torch26-cxx11-cu118-x86_64-linux": "sha256-55d13c115bd45757d78ed5fc44daeef82b36cd3d444d1b25c75c3d1206aef8e8",
+      "torch25-cxx98-cu118-x86_64-linux": "sha256-4d11fb99da76a1e7371e204446e956f3b2ea54e5400035fdfe1e7492534583a7",
+      "torch25-cxx11-cu124-x86_64-linux": "sha256-9b90d2d01d1ffc132a120fd2ea317994ee329b1b015d8c67f832f215b48a21ba",
+      "torch26-cxx11-cu124-x86_64-linux": "sha256-dd4a2add0d574032a6953da0648345e61438c953754522d6d08ff0327124b73a",
+      "torch26-cxx98-cu126-x86_64-linux": "sha256-4e172f5e5479fd94295e72e849cf73cdee25df4e16b0a2750e194de3002e7f48",
+      "torch25-cxx98-cu121-x86_64-linux": "sha256-02b4c801677dcc42c0223faba1c4f30fa0677077bf20b7f01cde80442e143c16",
+      "torch26-cxx98-cu118-x86_64-linux": "sha256-8d47863fe329cf7dc6d90c7305f9a288047b3af57a6fcd85a351ad84bd634fd7",
+      "torch26-cxx98-cu124-x86_64-linux": "sha256-783c6f8e8b5daf434b27b68911e65ae053ac6a71e7aef7c28f94510d3422277c",
+      "torch26-cxx11-cu126-x86_64-linux": "sha256-5314038ed6f34aafa7ab5d5c9bccd24b0ded1623fcd8ca889fb6f635ed2c4a20",
+      "torch25-cxx11-cu118-x86_64-linux": "sha256-81266aa4b1ae2a8dba6de510a9fba67c8aa4706e1b91ca2be999934f3f91f312"
+    }
+  }
+]

tests/hash_validation/pyproject.toml

@@ -0,0 +1,2 @@
+[tool.kernels.dependencies]
+"kernels-community/activation" = ">=0.0.2"

tests/test_hash_validation.py

@@ -0,0 +1,18 @@
+from dataclasses import dataclass
+from pathlib import Path
+
+from hf_kernels.cli import download_kernels
+
+
+# Mock download arguments class.
+@dataclass
+class DownloadArgs:
+    all_variants: bool
+    project_dir: Path
+
+
+def test_hash_validation():
+    project_dir = Path(__file__).parent / "hash_validation"
+
+    download_kernels(DownloadArgs(all_variants=False, project_dir=project_dir))
+    download_kernels(DownloadArgs(all_variants=True, project_dir=project_dir))