Ishaank18 commited on
Commit
875ba18
·
verified ·
1 Parent(s): c5b371e

Upload HSDN model

Browse files
Files changed (4) hide show
  1. README.md +40 -0
  2. best_model.pt +3 -0
  3. config.json +30 -0
  4. final_model.pt +3 -0
README.md ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # HSDN - Hierarchical Semantic Decomposition Network
2
+
3
+ A hierarchical retriever that decomposes embeddings into multi-resolution subspaces:
4
+ - **Level 1 (64D)**: Domain-level semantics
5
+ - **Level 2 (128D)**: Subdomain-level semantics
6
+ - **Level 3 (576D)**: Instance-level semantics
7
+
8
+ ## Usage
9
+
10
+ ```python
11
+ import torch
12
+ from sentence_transformers import SentenceTransformer
13
+
14
+ # Load base encoder
15
+ encoder = SentenceTransformer('BAAI/bge-base-en-v1.5')
16
+
17
+ # Load HSDN weights
18
+ from huggingface_hub import hf_hub_download
19
+ weights_path = hf_hub_download(repo_id="Ishaank18/hsdn-retriever", filename="best_model.pt")
20
+ state_dict = torch.load(weights_path, map_location='cpu')
21
+
22
+ # See cmrl_complete.py for full HSDNRetriever class
23
+ ```
24
+
25
+ ## Architecture
26
+
27
+ - Base encoder: BAAI/bge-base-en-v1.5
28
+ - Level 1 (Domain): 64D
29
+ - Level 2 (Subdomain): 128D
30
+ - Level 3 (Instance): 576D
31
+ - Total: 768D
32
+
33
+ ## Training
34
+
35
+ Trained with hierarchical contrastive loss + orthogonality constraints.
36
+
37
+ ## Part of CMRL Project
38
+
39
+ This retriever is part of the Contextual Multi-Resolution Learning (CMRL) system
40
+ for adaptive STEM tutoring.
best_model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f119c7834c002955cddd1bac5da116dd60805842f7c1eb6856467f6c9b752b5
3
+ size 450521185
config.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model": "BAAI/bge-base-en-v1.5",
3
+ "level1_dim": 64,
4
+ "level2_dim": 128,
5
+ "level3_dim": 576,
6
+ "num_domains": 6,
7
+ "num_subdomains": 20,
8
+ "subdomain_map": {
9
+ "Unknown": 0,
10
+ "World History Volume 2, from 1400": 1,
11
+ "Microbiology": 2,
12
+ "Chemistry 2e": 3,
13
+ "Elementary Algebra 2e": 4,
14
+ "Intermediate Algebra 2e": 5,
15
+ "Anatomy and Physiology 2e": 6,
16
+ "Psychology 2e": 7,
17
+ "Chemistry: Atoms First 2e": 8,
18
+ "Contemporary Mathematics": 9,
19
+ "University Physics Volume 2": 10,
20
+ "World History Volume 1, to 1500": 11,
21
+ "Makroekonomia - Podstawy": 12,
22
+ "Biology 2e": 13,
23
+ "University Physics Volume 1": 14,
24
+ "Workplace Software and Skills": 15,
25
+ "Astronomy 2e": 16,
26
+ "University Physics Volume 3": 17,
27
+ "Organic Chemistry": 18,
28
+ "Introduction to Sociology 3e": 19
29
+ }
30
+ }
final_model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a4bc78b2dfa0c22920421324656be534d369ea0b2d0ce7d2de4439c77a871b8
3
+ size 450521463