Add verified numeric caches with explicit provenance boundaries
Browse files- CACHES_MANIFEST.json +0 -0
- CACHE_SHA256SUMS +7 -0
- README.md +35 -2
- caches/cache_kary4_1024_expanded_source_seed424_v1.tar.gz +3 -0
- caches/cache_kary4_1024_validation3_strict_autoencoder_seed424_v2.tar.gz +3 -0
- caches/cache_kary4_leaf510_vs1024_v1_leaf1024.tar.gz +3 -0
- caches/cache_kary4_strict_autoencoder_official_test_v6.tar.gz +3 -0
- caches/cache_kary4_strict_autoencoder_v6_clean_development.tar.gz +3 -0
- caches/cache_kary4_strict_autoencoder_v6_clean_train.tar.gz +3 -0
CACHES_MANIFEST.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
CACHE_SHA256SUMS
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
18c52af51d37ee47119651667330129b0b613cbd0b4d0d1c2c132670b8583d66 caches/cache_kary4_1024_expanded_source_seed424_v1.tar.gz
|
| 2 |
+
837ef167422060465695f24153b0dabbbb2da480027e41c9a63d407a39fd8b95 caches/cache_kary4_1024_validation3_strict_autoencoder_seed424_v2.tar.gz
|
| 3 |
+
d1ee19b9eeb5651eba928e5e154298059b50b6164f3f39fee7ac7067296db0b1 caches/cache_kary4_leaf510_vs1024_v1_leaf1024.tar.gz
|
| 4 |
+
b615e66e3f9595d01e4fae0af0dfcb17dfe6a78490c8756982e6886ba7d3dc30 caches/cache_kary4_strict_autoencoder_official_test_v6.tar.gz
|
| 5 |
+
74d312bcb207a69f150dee4aa59c6620383cc9e974a3ae0fc4a6c348065f7ebe caches/cache_kary4_strict_autoencoder_v6_clean_development.tar.gz
|
| 6 |
+
726a62cce44022b3625402eae8c2e9a0efbb40d8e38fb0d0855f0ef9ee4a560a caches/cache_kary4_strict_autoencoder_v6_clean_train.tar.gz
|
| 7 |
+
1abaef15d7adb7e1b9f60ec7b0b4ffb9a727f85c42a63eb4ad89c8af3aebbb1f CACHES_MANIFEST.json
|
README.md
CHANGED
|
@@ -52,10 +52,43 @@ sha256sum --check SHA256SUMS
|
|
| 52 |
|
| 53 |
The manifest inside this model repository describes the payload at the revision containing it; the companion source/ZIP manifest records that enclosing immutable revision. This avoids a self-referential commit hash in a file that would itself change that hash. Checksums detect byte changes; they are not authenticity signatures.
|
| 54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
## Scientific and reproduction limits
|
| 56 |
|
| 57 |
-
- The released final clean-v6 models use **300 decoder-only updates plus 100 joint updates from the exact continuation-source checkpoint**, with a reinitialized decoder. They are not a from-scratch reconstruction.
|
| 58 |
-
- Restoring weights does not supply pinned teacher/corpus/task
|
| 59 |
- Separate older K-ary scaling checkpoints/probes and matched-pooling probes are outside the released 66-file inventory and remain unavailable unless explicitly listed in the companion access guide.
|
| 60 |
- The warm intermediates are not required to load the six final gradient-intervention outputs. Metadata-only copies do not satisfy guards requiring original serialized-byte hashes.
|
| 61 |
- Saved-score/statistical replay is not fresh model evaluation. Full-document roots have no direct full-root teacher target; retrieval results do not establish consistent superiority over pooling.
|
|
|
|
| 52 |
|
| 53 |
The manifest inside this model repository describes the payload at the revision containing it; the companion source/ZIP manifest records that enclosing immutable revision. This avoids a self-referential commit hash in a file that would itself change that hash. Checksums detect byte changes; they are not authenticity signatures.
|
| 54 |
|
| 55 |
+
## Separately versioned recovered caches
|
| 56 |
+
|
| 57 |
+
Six numeric cache bundles are an additive release, **not additions to or replacements for the immutable 66-checkpoint inventory**. `CACHES_MANIFEST.json` records all 8,075 extracted files, archive/member hashes and the two explicitly mapped metadata derivatives; `CACHE_SHA256SUMS` covers the six bundles and that manifest. The review source's `reproducibility/HOSTED_CACHES.json` pins their enclosing immutable revision separately from the weights.
|
| 58 |
+
|
| 59 |
+
| Cache directory | Role |
|
| 60 |
+
| --- | --- |
|
| 61 |
+
| `cache_kary4_strict_autoencoder_v6_clean_train` | Original clean-v6 TRAIN leaves/teacher targets; original manifest matches the preserved historical checksum. |
|
| 62 |
+
| `cache_kary4_strict_autoencoder_v6_clean_development` | Clean-v6 development safe/full leaves and full-document indexes. |
|
| 63 |
+
| `cache_kary4_strict_autoencoder_official_test_v6` | Official safe/full leaves; explicitly identified anonymous manifest/COMPLETE derivatives. |
|
| 64 |
+
| `cache_kary4_1024_expanded_source_seed424_v1` | Earlier expanded-source TRAIN/validation leaves and targets. |
|
| 65 |
+
| `cache_kary4_leaf510_vs1024_v1_leaf1024` | Earlier leaf1024 TRAIN/validation tensors and sidecars; no original root manifest/COMPLETE was recovered. |
|
| 66 |
+
| `cache_kary4_1024_validation3_strict_autoencoder_seed424_v2` | Separate validation3 leaves and targets. |
|
| 67 |
+
|
| 68 |
+
The bundles total **911,050,856 bytes compressed**, **1,089,336,987 bytes extracted**. All **8,032 PT files are unchanged originals**. The official manifest alone replaces an identifying implementation-parent Git label with `anonymous-source-snapshot-01`; its COMPLETE points to the derivative manifest. Original/new metadata hashes are explicit. All tensor, index and sidecar bytes are preserved. Archive timestamps, ownership and permissions are neutralized.
|
| 69 |
+
|
| 70 |
+
All 8,032 tensor files loaded with `weights_only=True` on CPU and contained finite floating tensors. One real row per cache directory passed the published final model's recursive reducer and strict decoder. Complete static pickle/ZIP metadata and CRC inspection found no other tested identity/credential markers or recognized raw-text/token-sequence fields. These are scoped integrity/loadability checks, not a historical metric rerun, perfect privacy proof or full reproduction.
|
| 71 |
+
|
| 72 |
+
**Do not bypass frozen guards.** The official derivative is internally consistent but cannot satisfy the original manifest hash. Original config paths/hashes and some provenance hashes also differ from the normalized review layout. Frozen results/configs/guards remain unchanged. Some other study-specific vectors, tokens and task caches are still absent.
|
| 73 |
+
|
| 74 |
+
### Upstream attribution and terms
|
| 75 |
+
|
| 76 |
+
These are numeric representations derived from the following datasets and pretrained teacher, not raw corpus text, token sequences or redistributed teacher weights. Dataset cards identify the licenses below; those labels do not settle every underlying web-page right or whether a particular derived representation triggers attribution/share-alike obligations. Retain applicable source terms; this project asserts no replacement blanket license.
|
| 77 |
+
|
| 78 |
+
| Upstream creators/resource | Pinned source | Stated terms |
|
| 79 |
+
| --- | --- | --- |
|
| 80 |
+
| Hugging Face FineWeb maintainers and source contributors | [FineWeb, sample-10BT](https://huggingface.co/datasets/HuggingFaceFW/fineweb/tree/9bb295ddab0e05d785b879661af7260fed5140fc) | [ODC-BY 1.0](https://opendatacommons.org/licenses/by/1-0/) |
|
| 81 |
+
| Hugging Face FineWeb-Edu maintainers and source contributors | [FineWeb-Edu, sample-10BT](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu/tree/87f09149ef4734204d70ed1d046ddc9ca3f2b8f9) | [ODC-BY 1.0](https://opendatacommons.org/licenses/by/1-0/) |
|
| 82 |
+
| OpenWebText maintainers and source contributors | [OpenWebText](https://huggingface.co/datasets/Skylion007/openwebtext/tree/79d93d786212f7344586290adb811d4ae6a1762c) | [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/) |
|
| 83 |
+
| Wikimedia/Wikipedia contributors | [Wikipedia, 20231101.en](https://huggingface.co/datasets/wikimedia/wikipedia/tree/b04c8d1ceb2f5cd4588862100d08de323dccfbaa) | [CC BY-SA 3.0](https://creativecommons.org/licenses/by-sa/3.0/), [GFDL](https://www.gnu.org/licenses/fdl-1.3.html), as stated by the pinned card |
|
| 84 |
+
| Alibaba Tongyi Lab | [gte-modernbert-base](https://huggingface.co/Alibaba-NLP/gte-modernbert-base/tree/e7f32e3c00f91d699e8c43b53106206bcc72bb22) | [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0) |
|
| 85 |
+
|
| 86 |
+
Document/source-index identifiers and content hashes remain intentionally available for provenance and can link cached representations to upstream documents. Embeddings can retain information about their inputs; no source-document privacy guarantee is made.
|
| 87 |
+
|
| 88 |
## Scientific and reproduction limits
|
| 89 |
|
| 90 |
+
- The released final clean-v6 models use **300 decoder-only updates plus 100 joint updates from the exact continuation-source checkpoint**, with a reinitialized decoder. They are not a from-scratch reconstruction. Recovered historical recipes/statuses trace the earlier source models through two encoder stages, but those earlier tensors and complete historical trainer/execution bindings remain missing; see the review source's partial lineage record.
|
| 91 |
+
- Restoring weights and the six recovered caches does not supply every pinned teacher/corpus/task input, study-specific token/vector cache, compatible software or GPU resource. Historical source-fingerprint discrepancies remain documented; no numerical result or old hash is rewritten to make a rerun pass.
|
| 92 |
- Separate older K-ary scaling checkpoints/probes and matched-pooling probes are outside the released 66-file inventory and remain unavailable unless explicitly listed in the companion access guide.
|
| 93 |
- The warm intermediates are not required to load the six final gradient-intervention outputs. Metadata-only copies do not satisfy guards requiring original serialized-byte hashes.
|
| 94 |
- Saved-score/statistical replay is not fresh model evaluation. Full-document roots have no direct full-root teacher target; retrieval results do not establish consistent superiority over pooling.
|
caches/cache_kary4_1024_expanded_source_seed424_v1.tar.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:18c52af51d37ee47119651667330129b0b613cbd0b4d0d1c2c132670b8583d66
|
| 3 |
+
size 118248753
|
caches/cache_kary4_1024_validation3_strict_autoencoder_seed424_v2.tar.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:837ef167422060465695f24153b0dabbbb2da480027e41c9a63d407a39fd8b95
|
| 3 |
+
size 30035926
|
caches/cache_kary4_leaf510_vs1024_v1_leaf1024.tar.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d1ee19b9eeb5651eba928e5e154298059b50b6164f3f39fee7ac7067296db0b1
|
| 3 |
+
size 337256271
|
caches/cache_kary4_strict_autoencoder_official_test_v6.tar.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b615e66e3f9595d01e4fae0af0dfcb17dfe6a78490c8756982e6886ba7d3dc30
|
| 3 |
+
size 57278505
|
caches/cache_kary4_strict_autoencoder_v6_clean_development.tar.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74d312bcb207a69f150dee4aa59c6620383cc9e974a3ae0fc4a6c348065f7ebe
|
| 3 |
+
size 57965594
|
caches/cache_kary4_strict_autoencoder_v6_clean_train.tar.gz
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:726a62cce44022b3625402eae8c2e9a0efbb40d8e38fb0d0855f0ef9ee4a560a
|
| 3 |
+
size 310265807
|