Upgrade to MiniLM-L12-v2: NDCG@10 0.1831 (+9.3% over L6)

1_Pooling/config.json

@@ -0,0 +1,10 @@
+{
+    "word_embedding_dimension": 384,
+    "pooling_mode_cls_token": false,
+    "pooling_mode_mean_tokens": true,
+    "pooling_mode_max_tokens": false,
+    "pooling_mode_mean_sqrt_len_tokens": false,
+    "pooling_mode_weightedmean_tokens": false,
+    "pooling_mode_lasttoken": false,
+    "include_prompt": true
+}

README.md

@@ -0,0 +1,102 @@
+---
+# For God so loved the world that he gave his only begotten Son,
+# that whoever believes in him should not perish but have eternal life. - John 3:16
+tags:
+- sentence-transformers
+- sentence-similarity
+- feature-extraction
+- bible
+- scripture
+- topical-search
+license: mit
+language: en
+base_model: sentence-transformers/all-MiniLM-L12-v2
+datasets:
+- LoveJesus/biblical-topical-dataset-chirho
+metrics:
+- ndcg_at_10
+- mrr_at_10
+- map_at_10
+model-index:
+- name: biblical-topical-search-chirho
+  results:
+  - task:
+      type: information-retrieval
+      name: Information Retrieval
+    metrics:
+    - name: NDCG@10
+      type: ndcg_at_10
+      value: 0.1831
+    - name: MRR@10
+      type: mrr_at_10
+      value: 0.1415
+    - name: MAP@10
+      type: map_at_10
+      value: 0.1415
+    - name: Accuracy@10
+      type: accuracy_at_10
+      value: 0.3188
+---
+
+# Biblical Topical Search - Chirho
+
+Semantic search model for finding biblical passages by topic. Fine-tuned on Nave's Topical Bible (30,000+ entries) and Treasury of Scripture Knowledge cross-references to retrieve relevant KJV verses for any theological query.
+
+## Model Details
+
+- **Base Model**: sentence-transformers/all-MiniLM-L12-v2 (33M params, 384-dim)
+- **Training Data**: 136,000 query-passage pairs from Nave's Topical Bible + TSK
+- **Loss**: MultipleNegativesRankingLoss
+- **Training**: 5 epochs, batch 16, lr 2e-5, cosine scheduler
+- **Hardware**: Apple M4 Pro (MPS), ~2 hours
+
+## Metrics (v2 - MiniLM-L12-v2 upgrade)
+
+| Metric | v1 (L6-v2) | v2 (L12-v2) | Change |
+|--------|-----------|-------------|--------|
+| **NDCG@10** | 0.1676 | **0.1831** | +9.3% |
+| Accuracy@10 | 0.2882 | **0.3188** | +10.6% |
+| MRR@10 | 0.1305 | **0.1415** | +8.4% |
+| MAP@10 | 0.1305 | **0.1415** | +8.4% |
+
+### Training Trajectory (v2)
+
+| Epoch | NDCG@10 | Accuracy@10 |
+|-------|---------|-------------|
+| 1 | 0.1514 | 0.257 |
+| 2 | 0.1654 | 0.286 |
+| 3 | 0.1749 | 0.302 |
+| 4 | 0.1795 | 0.313 |
+| 5 | **0.1831** | **0.319** |
+
+## Usage
+
+```python
+from sentence_transformers import SentenceTransformer
+import numpy as np
+
+model = SentenceTransformer("LoveJesus/biblical-topical-search-chirho")
+
+query = "What does the Bible say about forgiveness?"
+verses = [
+    "For if ye forgive men their trespasses, your heavenly Father will also forgive you. - Matthew 6:14",
+    "In the beginning God created the heaven and the earth. - Genesis 1:1",
+    "As far as the east is from the west, so far hath he removed our transgressions from us. - Psalm 103:12",
+]
+
+query_emb = model.encode([query])
+verse_embs = model.encode(verses)
+
+scores = np.dot(verse_embs, query_emb.T).flatten()
+ranked = sorted(zip(scores, verses), reverse=True)
+for score, verse in ranked:
+    print(f"  {score:.3f}: {verse}")
+```
+
+## Part of Bible.Systems
+
+This model is part of the [Bible.Systems](https://bible.systems) AI models project — open-source machine learning for biblical scholarship.
+
+- **Live Demo**: [HuggingFace Space](https://huggingface.co/spaces/LoveJesus/biblical-topical-search-chirho)
+- **Dataset**: [LoveJesus/biblical-topical-dataset-chirho](https://huggingface.co/datasets/LoveJesus/biblical-topical-dataset-chirho)
+- **All Models**: [LoveJesus on HuggingFace](https://huggingface.co/LoveJesus)

config.json

@@ -0,0 +1,30 @@
+{
+  "add_cross_attention": false,
+  "architectures": [
+    "BertModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": null,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "eos_token_id": null,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 384,
+  "initializer_range": 0.02,
+  "intermediate_size": 1536,
+  "is_decoder": false,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "tie_word_embeddings": true,
+  "transformers_version": "5.1.0",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522
+}

config_sentence_transformers.json

@@ -0,0 +1,14 @@
+{
+  "__version__": {
+    "sentence_transformers": "5.2.2",
+    "transformers": "5.1.0",
+    "pytorch": "2.10.0"
+  },
+  "model_type": "SentenceTransformer",
+  "prompts": {
+    "query": "",
+    "document": ""
+  },
+  "default_prompt_name": null,
+  "similarity_fn_name": "cosine"
+}

eval/Information-Retrieval_evaluation_val-topical-chirho_results.csv

@@ -0,0 +1,11 @@
+epoch,steps,cosine-Accuracy@1,cosine-Accuracy@3,cosine-Accuracy@5,cosine-Accuracy@10,cosine-Precision@1,cosine-Recall@1,cosine-Precision@3,cosine-Recall@3,cosine-Precision@5,cosine-Recall@5,cosine-Precision@10,cosine-Recall@10,cosine-MRR@10,cosine-NDCG@10,cosine-MAP@10
+1.0,2125,0.063,0.1414,0.1826,0.2472,0.063,0.063,0.04713333333333333,0.1414,0.036520000000000004,0.1826,0.02472,0.2472,0.11534984126984148,0.14651173930780936,0.11534984126984127
+2.0,4250,0.0686,0.1512,0.1966,0.2656,0.0686,0.0686,0.05039999999999999,0.1512,0.03932000000000001,0.1966,0.026560000000000004,0.2656,0.12370682539682545,0.1572021694578249,0.12370682539682538
+3.0,6375,0.0682,0.1584,0.2042,0.2782,0.0682,0.0682,0.0528,0.1584,0.040839999999999994,0.2042,0.027820000000000004,0.2782,0.12690301587301606,0.16260026203280784,0.12690301587301586
+4.0,8500,0.0696,0.1594,0.2062,0.2846,0.0696,0.0696,0.053133333333333324,0.1594,0.04124,0.2062,0.028460000000000003,0.2846,0.12954928571428573,0.16605838585267033,0.1295492857142857
+5.0,10625,0.0702,0.1584,0.2042,0.2882,0.0702,0.0702,0.0528,0.1584,0.04084,0.2042,0.028820000000000005,0.2882,0.1304676984126985,0.16755473492351874,0.1304676984126984
+1.0,8500,0.064,0.1458,0.1926,0.257,0.064,0.064,0.0486,0.1458,0.03852,0.1926,0.0257,0.257,0.11865857142857139,0.15137766715248815,0.11865857142857145
+2.0,17000,0.067,0.1586,0.206,0.286,0.067,0.067,0.052866666666666666,0.1586,0.0412,0.206,0.0286,0.286,0.12828412698412714,0.16544290848995974,0.128284126984127
+3.0,25500,0.073,0.1676,0.2158,0.3016,0.073,0.073,0.05586666666666666,0.1676,0.043160000000000004,0.2158,0.030160000000000003,0.3016,0.13592452380952405,0.17488514026229263,0.13592452380952383
+4.0,34000,0.0726,0.1706,0.2236,0.3128,0.0726,0.0726,0.05686666666666666,0.1706,0.04472,0.2236,0.03128000000000001,0.3128,0.1384629365079366,0.17946596649694194,0.1384629365079365
+5.0,42500,0.0764,0.1708,0.2234,0.3188,0.0764,0.0764,0.05693333333333333,0.1708,0.044680000000000004,0.2234,0.03188,0.3188,0.14151626984126994,0.18311836451970726,0.14151626984126986

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ee032913bf1067aeb6d04e9e0734fc2edf20594cd275a805b56cc867376a94d
+size 133462104

modules.json

@@ -0,0 +1,20 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.models.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.models.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.models.Normalize"
+  }
+]

sentence_bert_config.json

@@ -0,0 +1,4 @@
+{
+    "max_seq_length": 128,
+    "do_lower_case": false
+}

tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "backend": "tokenizers",
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": true,
+  "is_local": false,
+  "mask_token": "[MASK]",
+  "model_max_length": 128,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}