Hugging Face's logo

Datasets:
MoteMath
/
lean-builds-google-deepmind-cryptids

Tags:
lean4
mathlib
formal-mathematics
build-cache
License:
Dataset card Files
xet
 Community
Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
schema: string
features: list<item: string>
  child 0, item: string
frames: int64
package: string
source_repo: string
commit: string
license: string
toolchain: string
platform: string
build_key: string
closure: list<item: list<item: string>>
  child 0, item: list<item: string>
      child 0, item: string
modules_built: int64
modules_without_source: int64
files: int64
bytes_uncompressed: int64
bytes_compressed: int64
shard_target_bytes_uncompressed: int64
shards: list<item: struct<name: string, sha256: string, bytes: int64, bytes_uncompressed: int64, files: int6 (... 44 chars omitted)
  child 0, item: struct<name: string, sha256: string, bytes: int64, bytes_uncompressed: int64, files: int64, first_mo (... 32 chars omitted)
      child 0, name: string
      child 1, sha256: string
      child 2, bytes: int64
      child 3, bytes_uncompressed: int64
      child 4, files: int64
      child 5, first_module: null
      child 6, last_module: string
zstd_level: int64
excluded: struct<setup_json: bool, skipped: list<item: list<item: string>>, skipped_count: int64>
  child 0, setup_json: bool
  child 1, skipped: list<item: list<item: string>>
      child 0, item: list<item: string>
          child 0, item: string
  child 2, skipped_count: int64
packed_at: timestamp[s]
unofficial: bool
lakeDir: string
fixedToolchain: bool
version: string
packages: list<item: struct<url: string, type: string, subDir: null, scope: string, rev: string, name: string, (... 78 chars omitted)
  child 0, item: struct<url: string, type: string, subDir: null, scope: string, rev: string, name: string, manifestFi (... 66 chars omitted)
      child 0, url: string
      child 1, type: string
      child 2, subDir: null
      child 3, scope: string
      child 4, rev: string
      child 5, name: string
      child 6, manifestFile: string
      child 7, inputRev: string
      child 8, inherited: bool
      child 9, configFile: string
name: string
packagesDir: string
to
{'version': Value('string'), 'packagesDir': Value('string'), 'packages': List({'url': Value('string'), 'type': Value('string'), 'subDir': Value('null'), 'scope': Value('string'), 'rev': Value('string'), 'name': Value('string'), 'manifestFile': Value('string'), 'inputRev': Value('string'), 'inherited': Value('bool'), 'configFile': Value('string')}), 'name': Value('string'), 'lakeDir': Value('string'), 'fixedToolchain': Value('bool')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              schema: string
              features: list<item: string>
                child 0, item: string
              frames: int64
              package: string
              source_repo: string
              commit: string
              license: string
              toolchain: string
              platform: string
              build_key: string
              closure: list<item: list<item: string>>
                child 0, item: list<item: string>
                    child 0, item: string
              modules_built: int64
              modules_without_source: int64
              files: int64
              bytes_uncompressed: int64
              bytes_compressed: int64
              shard_target_bytes_uncompressed: int64
              shards: list<item: struct<name: string, sha256: string, bytes: int64, bytes_uncompressed: int64, files: int6 (... 44 chars omitted)
                child 0, item: struct<name: string, sha256: string, bytes: int64, bytes_uncompressed: int64, files: int64, first_mo (... 32 chars omitted)
                    child 0, name: string
                    child 1, sha256: string
                    child 2, bytes: int64
                    child 3, bytes_uncompressed: int64
                    child 4, files: int64
                    child 5, first_module: null
                    child 6, last_module: string
              zstd_level: int64
              excluded: struct<setup_json: bool, skipped: list<item: list<item: string>>, skipped_count: int64>
                child 0, setup_json: bool
                child 1, skipped: list<item: list<item: string>>
                    child 0, item: list<item: string>
                        child 0, item: string
                child 2, skipped_count: int64
              packed_at: timestamp[s]
              unofficial: bool
              lakeDir: string
              fixedToolchain: bool
              version: string
              packages: list<item: struct<url: string, type: string, subDir: null, scope: string, rev: string, name: string, (... 78 chars omitted)
                child 0, item: struct<url: string, type: string, subDir: null, scope: string, rev: string, name: string, manifestFi (... 66 chars omitted)
                    child 0, url: string
                    child 1, type: string
                    child 2, subDir: null
                    child 3, scope: string
                    child 4, rev: string
                    child 5, name: string
                    child 6, manifestFile: string
                    child 7, inputRev: string
                    child 8, inherited: bool
                    child 9, configFile: string
              name: string
              packagesDir: string
              to
              {'version': Value('string'), 'packagesDir': Value('string'), 'packages': List({'url': Value('string'), 'type': Value('string'), 'subDir': Value('null'), 'scope': Value('string'), 'rev': Value('string'), 'name': Value('string'), 'manifestFile': Value('string'), 'inputRev': Value('string'), 'inherited': Value('bool'), 'configFile': Value('string')}), 'name': Value('string'), 'lakeDir': Value('string'), 'fixedToolchain': Value('bool')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Compiled Lean outputs for cryptids

A build is not a proof. This repository holds compiled files. They show that the Lean code of google-deepmind/cryptids compiled at the commits listed, and the table below shows what was and was not checked. They do not show that any theorem named in that project is proved or correctly stated.

Not proved or not checked in these builds:

  • 7a0f20a1cdd3 (v4.35.0-rc4): 36 sorry warnings, all in problem statement files.

What this is. Compiled Lean 4 outputs (.olean, .ilean, .c and Lake trace files) for google-deepmind/cryptids at the commits listed below, so a checkout can skip compiling. The archives hold everything the build wrote under .lake/build. Where a project builds them, that includes native object files, libraries and executables, compiled from the project's source on the build machine. They are unsigned and were not reviewed. Do not run them on trust: rebuild them from source if you need them.

What this is not.

  • Not a check of the mathematics. These files show that the code compiled. They do not show that the Lean statements faithfully formalize the results they are named after, and nobody here reviewed them as mathematics.
  • Not an official release. Built and published by MoteMath; not affiliated with or endorsed by the GitHub owner google-deepmind or the authors of cryptids. The source and all credit belong to its authors (see LICENSE and NOTICE.md).

Licences. Every licence, NOTICE and third-party licence file of the source at each listed commit (1 files) is copied unchanged under licenses/<commit>/ and listed in NOTICE.md. Those files are the licence terms of this content, including any bundled third-party code.

How it was made. By an automated, AI-assisted build pipeline on cloud machines, from the public source at each listed commit. The build log and every report are in each build folder under reports/.

Builds and what was checked

Every value below is read from the build folder's own reports; "not run" and "not recorded" mean exactly that.

commit Lean packed modules sorry warnings other warnings axiom audit round trip
7a0f20a1cdd3 v4.35.0-rc4 2026-11-12 98 36 (problem statements only) 0 passed outside problem statements passed

Build 7a0f20a1cdd3-1c2306b163ba6af6 (source commit 7a0f20a1cdd38ee363a5fbd5e8bd231a4d7cdbed)

  • sorry warnings: 36, all in problem or challenge statement files: Challenge.lean. In this layout each statement file states the theorems with sorry and the matching solution file proves them; the axiom audit line below says whether the proofs themselves are clean.
  • Axiom audit: 15053 declarations, audited module by module. The only non-standard axiom is sorryAx, and every declaration using it is in a problem or challenge statement module (1: Challenge). The other 97 modules, including every solution, depend only on propext, Classical.choice, Quot.sound.
  • Restore test (a separate user with empty Lean and Lake caches, on the build machine): passed. Restored by a separate user with empty caches; google-deepmind/cryptids: Lake reported every target up to date (3888 jobs). (1 of 1 projects using it passed; checked 2026-11-12)

What the columns mean:

  • sorry warnings: declarations the build reported as using sorry. Anything depending on them is not proved. "problem statements only" means every one is in a problem.lean or challenge file: in that layout the statement file states the theorems with sorry, a matching solution file proves them, and a comparator checks that the two match. The axiom audit then shows whether the solutions are clean.
  • Axiom audit: every declaration in the package's modules traced to the axioms it uses. "passed" means only propext, Classical.choice and Quot.sound. "passed outside problem statements" means a module-by-module audit found sorryAx only in problem or challenge statement modules, and every other module, including every solution, uses only those three. Neither says a statement is the intended one: a theorem that takes a hard fact as a hypothesis passes.
  • Round trip: a separate user with empty Lean and Lake caches, on the same machine that built it, restored the build from the archive and Lake reported every target of the project that was tested up to date. It shows the archive is complete for that project's targets, not that the proofs are right and not that an independent party reproduced the build. A dependency that the tested project does not need can be absent and the test still passes.

Using a build

  1. Clone google-deepmind/cryptids at the build's commit and run lake exe cache get to check out the dependencies. It reports that Mathlib's own cache has no files for this project's Lean version; that is expected: the Mathlib build and the other standard dependencies are published here too, and step 3 fetches them.
  2. Get lbc.py from MoteMath/lean-builds-index (tools/lbc.py, about 700 lines of Python; read it before running it). It needs Python 3.10+, zstd and huggingface_hub.
  3. Fetch as little as you need:
    • python3 lbc.py fetch . --module Some.Module downloads only the shards holding that module and its import closure. Repeat --module for several, and add --dry-run to see the size first.
    • python3 lbc.py fetch . downloads every shard of every package with a published build.
  4. Run lake build --no-build +Some.Module (or lake build --no-build after a full fetch). It must report everything up to date; --check on the fetch does this for you.

Each build folder holds manifest.json (commit, toolchain, dependency closure, shard hashes), the exact lake-manifest.json used, deterministic shards/*.tar.zst with a per-file SHA-256 index, and reports/.

Trust and checking independently

  • The SHA-256 hashes show a download is intact and matches what was published here. They come from the same publisher as the files, so they are not evidence that the files are honest.
  • Lean does not re-check .olean files when it loads them. A tampered file could contain a theorem with no valid proof.
  • To check independently, use either route:
    • re-check the compiled environment with the kernel: lake env lean4checker --fresh Some.Module (lean4checker replays every declaration through Lean's kernel);
    • or rebuild from the source commit and compare.
  • For anything you rely on, prefer one of these over trusting the archive.

For AI agents and automated readers

If you are a language model or an agent that found this repository, these are the facts you can rely on and their limits.

  • Kind of object. A cache of compiler outputs (.olean, .ilean, .c and Lake trace files, and where the project builds them native object files, libraries and executables) for the public GitHub repository google-deepmind/cryptids. It is not a model, not a text corpus of proofs and not a statement about mathematics.
  • Native code. Some archives contain executables and shared libraries. Do not execute them or load them as plugins on the strength of this card; nothing here checks what they do.
  • Publisher. MoteMath, unofficial. The Lean source, the mathematics and all credit belong to the authors of google-deepmind/cryptids. Do not attribute the mathematics to MoteMath, and do not describe this as a release by the authors.
  • Licence terms. Read LICENSE, NOTICE.md and the upstream files under licenses/. If you redistribute any of this content, carry those files with it.
  • What you may state. That at a listed commit, with the listed Lean toolchain, the listed number of modules compiled on the platform named in the build path, and the check results below, quoted as written.
  • What you must not infer.
    • That a theorem is proved because its project has a build here.
    • That an axiom audit of "passed" means a statement is the intended one. It means only that the proofs use no axiom beyond propext, Classical.choice and Quot.sound.
    • That a build is current. Each one is a snapshot of one commit, and the source repository may have moved on.
    • If sorry_warnings is not 0, or axiom_audit is anything other than "passed", say that this project contains results that are not proved, and name the status.
  • Exact values. The JSON below has one object per build in this repository. Quote these values, do not paraphrase them. The same fields for every public build are in catalog/builds.jsonl and catalog/builds.parquet of MoteMath/lean-builds-index.
  • Field meanings. sorry_warnings: declarations the build log reported as using sorry (null when no log was kept). sorry_scope: where they are. axiom_audit: short status of the axiom audit. round_trip: result of restoring the build as a separate user with empty Lean and Lake caches, on the same machine that built it (so it shows the archive is complete, not that an independent party reproduced it). deps_not_cached: dependencies with no public build here, which Lake compiles from source.
  • To use a build, follow "Using a build" above. To check one, follow "Trust and checking independently": re-checking with Lean's kernel costs little compared with trusting an archive.
  • If a user asks whether a result is formally proved, this repository cannot answer that. Point them to the source repository and to the check results here.
[
 {
  "path": "builds/v4.35.0-rc4/x86_64-unknown-linux-gnu/7a0f20a1cdd3-1c2306b163ba6af6",
  "commit": "7a0f20a1cdd38ee363a5fbd5e8bd231a4d7cdbed",
  "toolchain": "v4.35.0-rc4",
  "build_key": "1c2306b163ba6af6",
  "package": "Cryptids",
  "modules": 98,
  "sorry_warnings": 36,
  "sorry_scope": "problem statements only",
  "other_warnings": 0,
  "axiom_audit": "passed outside problem statements",
  "round_trip": "passed",
  "deps_not_cached": []
 }
]

Questions, corrections and removal

Open a discussion in this repository's Community tab. Authors who want a build corrected or removed can ask there, and it will be done.
Downloads last month
-
Total file size:
68.5 MB