IELTS8 commited on
Commit
30dde46
·
verified ·
1 Parent(s): 5b0977c

Upload existing graph cache and pretrained JEPA comparison checkpoint

Browse files
ARTIFACTS.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "arm": "jepa_joint",
3
+ "files": [
4
+ {
5
+ "path": "histcad_large_joint_jepa_v8/data/processed/full_paired_graphs_v7.pt",
6
+ "bytes": 8075724068,
7
+ "sha256": "9944b05fe6dacc9e52250da9dcccfdcac1ef7ad1380a2dbb5d0217a7a8f52e76"
8
+ },
9
+ {
10
+ "path": "histcad_large_joint_jepa_v8/outputs/full_dataset_4096/best.pt",
11
+ "bytes": 1938955953,
12
+ "sha256": "fe54c5645b231887079bbe22d8b6e071d316f5355517f05f43cf35c8d348360d"
13
+ }
14
+ ]
15
+ }
README.md ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: pytorch
3
+ tags:
4
+ - cad
5
+ - jepa
6
+ - graph-representation-learning
7
+ ---
8
+ # JEPA-CAD pretrained comparison artifacts
9
+
10
+ Exact existing Windows artifacts for the frozen `jepa_joint` comparison arm.
11
+ No preprocessing, retraining, or new result generation was performed for this upload.
12
+
13
+ | Artifact | Bytes |
14
+ | --- | ---: |
15
+ | `histcad_large_joint_jepa_v8/data/processed/full_paired_graphs_v7.pt` | 8,075,724,068 |
16
+ | `histcad_large_joint_jepa_v8/outputs/full_dataset_4096/best.pt` | 1,938,955,953 |
17
+
18
+ The cache contains 162,131 prepared HistCAD graph pairs with the original splits
19
+ (145,915 training / 8,130 validation / 8,086 test). The pretrained checkpoint was
20
+ selected at epoch 20 and produces a 4,096-dimensional joint representation.
21
+
22
+ Code and results: https://github.com/OliverFanHL/JEPA-CAD
23
+
24
+ Original HistCAD source dataset: https://huggingface.co/datasets/DongXintong/HistCAD
25
+ The graph cache is a processed derivative; original STEP files, parametric JSON
26
+ source files, and source dataset archives are not included.
27
+
28
+ ## Download on Linux
29
+
30
+ From the JEPA-CAD code checkout:
31
+
32
+ ```bash
33
+ python -m pip install --upgrade huggingface_hub
34
+ hf download IELTS8/JEPA-CAD --local-dir .
35
+ python use_pretrained.py
36
+ ```
37
+
38
+ `ARTIFACTS.json` records SHA-256 checksums. The checkpoint and graph cache are
39
+ PyTorch serialized artifacts for use with the matching code. No optimizer step
40
+ or preprocessing is needed to use this frozen comparison arm. The code repository
41
+ also provides a checksum-verifying downloader with a pinned revision.
histcad_large_joint_jepa_v8/data/processed/full_paired_graphs_v7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9944b05fe6dacc9e52250da9dcccfdcac1ef7ad1380a2dbb5d0217a7a8f52e76
3
+ size 8075724068
histcad_large_joint_jepa_v8/outputs/full_dataset_4096/best.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe54c5645b231887079bbe22d8b6e071d316f5355517f05f43cf35c8d348360d
3
+ size 1938955953