abanwild commited on
Commit
80491d5
·
verified ·
1 Parent(s): 77dbe05

Add comprehensive dataset documentation

Browse files
Files changed (1) hide show
  1. README.md +42 -20
README.md CHANGED
@@ -1,32 +1,54 @@
1
- # PeatLearn Embeddings Dataset
 
 
 
2
 
3
- This dataset contains pre-computed embeddings for the Ray Peat bioenergetic corpus used in the PeatLearn AI platform.
4
 
5
- ## Files
6
 
7
- - `embeddings_20250728_221826.npy` - NumPy array of embeddings (310MB, latest complete run)
8
- - `metadata_20250728_221826.json` - Corresponding metadata and text chunks (21MB)
9
- - `ray_peat_embeddings_20250728_221825.pkl` - Pickled embeddings in alternative format (330MB)
10
- - `checkpoint.json` - Processing checkpoint information
11
- - `embedding_report.json` - Summary statistics and processing report
 
 
 
 
 
 
 
 
 
 
12
 
13
  ## Usage
14
 
 
15
  ```python
16
- # Download embeddings using the PeatLearn scripts
17
- python embedding/download_from_hf.py --repo abanwild/peatlearn-embeddings --dest embedding/vectors
 
18
 
19
- # Or using Hugging Face CLI
20
- hf download abanwild/peatlearn-embeddings --repo-type=dataset --local-dir embedding/vectors
21
- ```
 
 
22
 
23
- ## Model Information
 
24
 
25
- - **Embedding Model**: text-embedding-004 (Google)
26
- - **Dimensions**: 768
27
- - **Total Documents**: 552 Ray Peat documents
28
- - **Processing Date**: July 28, 2025
 
 
29
 
30
- ## Repository
 
31
 
32
- Source code: https://github.com/thewildofficial/PeatLearn
 
 
1
+ ---
2
+ title: Ray Peat Learning Embeddings
3
+ description: Dense vector embeddings for Ray Peat bioenergetic corpus, generated using sentence-transformers/all-mpnet-base-v2
4
+ ---
5
 
6
+ # Ray Peat Learning Embeddings
7
 
8
+ This dataset contains dense vector embeddings for the Ray Peat bioenergetic corpus, designed for semantic search and retrieval-augmented generation (RAG) applications.
9
 
10
+ ## Dataset Contents
11
+
12
+ - **embeddings_20250728_221826.npy**: Dense vector embeddings (768-dimensional)
13
+ - **metadata_20250728_221826.json**: Metadata and text content for each embedding
14
+ - **ray_peat_embeddings_20250728_221825.pkl**: Complete embeddings in pickle format
15
+ - **checkpoint.json**: Processing checkpoint information
16
+ - **embedding_report.json**: Generation statistics and metrics
17
+
18
+ ## Technical Details
19
+
20
+ - **Embedding Model**: sentence-transformers/all-mpnet-base-v2
21
+ - **Dimensions**: 768
22
+ - **Generation Date**: July 28, 2025
23
+ - **Total Embeddings**: ~18,500+ text segments
24
+ - **Source**: Ray Peat bioenergetic articles and materials
25
 
26
  ## Usage
27
 
28
+ ### Download Embeddings
29
  ```python
30
+ from huggingface_hub import snapshot_download
31
+ import numpy as np
32
+ import json
33
 
34
+ # Download all files
35
+ local_dir = snapshot_download(
36
+ repo_id="abanwild/peatlearn-embeddings",
37
+ repo_type="dataset"
38
+ )
39
 
40
+ # Load embeddings
41
+ embeddings = np.load(f"{local_dir}/embeddings_20250728_221826.npy")
42
 
43
+ # Load metadata
44
+ with open(f"{local_dir}/metadata_20250728_221826.json", "r") as f:
45
+ metadata = json.load(f)
46
+
47
+ print(f"Loaded {len(embeddings)} embeddings with {embeddings.shape[1]} dimensions")
48
+ ```
49
 
50
+ ### Integration with PeatLearn
51
+ This dataset is designed to work with the [PeatLearn](https://github.com/thewildofficial/PeatLearn) project. The embeddings enable semantic search across Ray Peat's bioenergetic research.
52
 
53
+ ## License
54
+ This dataset is released under the same license as the PeatLearn project.