Datasets:
Initial corpus upload from rewriter-probe
Browse files- README.md +18 -0
- captions.json +0 -0
- embeddings.npy +3 -0
README.md
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
language:
|
| 4 |
+
- multilingual
|
| 5 |
+
tags:
|
| 6 |
+
- text-to-motion
|
| 7 |
+
- humanml3d
|
| 8 |
+
- rewriter
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# AnimoFlow Rewriter Corpus
|
| 12 |
+
|
| 13 |
+
Pre-extracted + pre-embedded HumanML3D caption corpus, used by AnimoFlows multilingual prompt rewriter (Qwen2.5-1.5B-Instruct + RAFSL few-shot template).
|
| 14 |
+
|
| 15 |
+
- captions.json: 52,839 unique English motion captions extracted from EricGuo5513/HumanML3D
|
| 16 |
+
- embeddings.npy: (52839, 384) float16 — embedded once with sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
|
| 17 |
+
|
| 18 |
+
Baked into the AnimoFlow demo Space at build time. See vault://wiki/concepts/Prompt rewriter front-end plan for context.
|
captions.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
embeddings.npy
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f5c62430bb3ce7d67d3b997698120e02ddb2dc138fc7cc786902072568fca912
|
| 3 |
+
size 40580480
|