Document LFTC dataset and checkpoint snapshot
Browse files
README.md
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
- zh
|
| 5 |
+
license: other
|
| 6 |
+
license_name: source-specific-licenses
|
| 7 |
+
license_link: https://github.com/eyanhe/NLP-Course-Project-LFTC/blob/main/assets.lock.json
|
| 8 |
+
task_categories:
|
| 9 |
+
- text-classification
|
| 10 |
+
tags:
|
| 11 |
+
- checkpoints
|
| 12 |
+
- reproducibility
|
| 13 |
+
---
|
| 14 |
+
# LFTC: datasets and checkpoints
|
| 15 |
+
|
| 16 |
+
Upload in progress; use the verified manifest before relying on completeness.
|
| 17 |
+
|
| 18 |
+
This is the large-asset companion to [NLP-Course-Project-LFTC](https://github.com/eyanhe/NLP-Course-Project-LFTC).
|
| 19 |
+
Code, reports, figures, PowerPoint, Excel, metrics and per-example predictions are in GitHub.
|
| 20 |
+
|
| 21 |
+
The local snapshot contains **2,044 asset files**, **56,530,278,162 bytes**.
|
| 22 |
+
Files under `artifacts/` retain their original project-relative paths, including datasets, prepared inputs, pretrained models, final/local checkpoints, feature bundles, and historical candidate checkpoints.
|
| 23 |
+
[artifact-manifest.json](artifact-manifest.json) records paths, sizes, SHA256 values and verification status.
|
| 24 |
+
|
| 25 |
+
## Restore
|
| 26 |
+
|
| 27 |
+
```sh
|
| 28 |
+
hf download Eyanhe/LFTC_Dataset_And_Checkpoints --repo-type dataset --local-dir ../lftc-assets
|
| 29 |
+
```
|
| 30 |
+
|
| 31 |
+
Merge the contents of the downloaded `artifacts/` directory into the GitHub project root, preserving all relative paths, including hidden candidate directories.
|
| 32 |
+
The snapshot covers files actually present on the local workstation; it does not claim that every checkpoint produced on the training cluster has been downloaded.
|
| 33 |
+
|
| 34 |
+
## Provenance and scope
|
| 35 |
+
|
| 36 |
+
The study has 428 completed v3 records, including 232 current-research records, 226 with V metrics and 42 fixed models with formal Q evaluations.
|
| 37 |
+
These counts describe experiments, not a promise of 428 complete checkpoint folders. The original 2,859-item matrix was not fully executed.
|
| 38 |
+
|
| 39 |
+
Each original dataset and pretrained model retains its own source-specific license. The mixed collection is not relicensed under one blanket license.
|
| 40 |
+
Original asset locks and model/readme metadata are retained at their corresponding paths and in GitHub.
|
| 41 |
+
Local credentials, runtimes, private cluster-login helpers, regenerable MinHash work cache and duplicate transfer archives are outside this publication.
|