Record ModelScope mirror receipts (R11)
Browse files- CHANGELOG.md +14 -0
CHANGELOG.md
CHANGED
|
@@ -247,3 +247,17 @@ release is uploaded, and every accepted result below was produced with
|
|
| 247 |
`acquisition_calls=0`) and the 28-test suite from the cloned `tests/`
|
| 248 |
passes against that install.
|
| 249 |
- The model-package channel remains Hugging Face; no weights on GitHub.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 247 |
`acquisition_calls=0`) and the 28-test suite from the cloned `tests/`
|
| 248 |
passes against that install.
|
| 249 |
- The model-package channel remains Hugging Face; no weights on GitHub.
|
| 250 |
+
|
| 251 |
+
### ModelScope mirror (same day, owner-approved)
|
| 252 |
+
|
| 253 |
+
- Public mirror repo created at
|
| 254 |
+
[`kialajin/cognitive-learning-lab`](https://modelscope.cn/models/kialajin/cognitive-learning-lab)
|
| 255 |
+
(token validated via SDK `whoami`, account `kialajin`).
|
| 256 |
+
- 14 files uploaded **byte-identical** to the HF repo (thin package, no
|
| 257 |
+
weights, wheel via LFS) — commit `d0aaacbffe…`.
|
| 258 |
+
- Download-back verification: 10/10 flat-layout doc entries match
|
| 259 |
+
byte-for-byte; wheel/sdist hashes match the current build
|
| 260 |
+
(`64e02c1a…`/`d69ad3ce…`). The unmatched `SHA256SUMS` lines are the flat
|
| 261 |
+
working-tree layout, not missing content.
|
| 262 |
+
- Channels: Hugging Face primary, ModelScope mirror, GitHub source copy;
|
| 263 |
+
all three carry the same artifact identity.
|