Add new SentenceTransformer model

1_Pooling/config.json

@@ -1,7 +1,7 @@
 {
-    "word_embedding_dimension": 768,
-    "pooling_mode_cls_token": true,
-    "pooling_mode_mean_tokens": false,
+    "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,

README.md

@@ -7,7 +7,7 @@ tags:
 - generated_from_trainer
 - dataset_size:90000
 - loss:MultipleNegativesRankingLoss
-base_model: Alibaba-NLP/gte-modernbert-base
+base_model: thenlper/gte-small
 widget:
 - source_sentence: what is the maximum i can contribute to a traditional ira
   sentences:
@@ -121,7 +121,7 @@ metrics:
 - cosine_mrr@10
 - cosine_map@100
 model-index:
-- name: SentenceTransformer based on Alibaba-NLP/gte-modernbert-base
+- name: SentenceTransformer based on thenlper/gte-small
   results:
   - task:
       type: information-retrieval
@@ -131,49 +131,49 @@ model-index:
       type: NanoMSMARCO
     metrics:
     - type: cosine_accuracy@1
-      value: 0.46
+      value: 0.32
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.64
+      value: 0.56
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.76
+      value: 0.66
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.84
+      value: 0.76
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.46
+      value: 0.32
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.21333333333333332
+      value: 0.18666666666666668
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.15200000000000002
+      value: 0.132
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.08399999999999999
+      value: 0.07600000000000001
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.46
+      value: 0.32
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.64
+      value: 0.56
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.76
+      value: 0.66
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.84
+      value: 0.76
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.6415212118347274
+      value: 0.543482168903518
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.5786904761904762
+      value: 0.474047619047619
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.586056624889745
+      value: 0.48355956186670473
       name: Cosine Map@100
   - task:
       type: information-retrieval
@@ -189,43 +189,43 @@ model-index:
       value: 0.56
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.62
+      value: 0.6
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.74
+      value: 0.76
       name: Cosine Accuracy@10
     - type: cosine_precision@1
       value: 0.42
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.19333333333333333
+      value: 0.2
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.128
+      value: 0.132
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.07800000000000001
+      value: 0.08199999999999999
       name: Cosine Precision@10
     - type: cosine_recall@1
       value: 0.39
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.53
+      value: 0.55
       name: Cosine Recall@3
     - type: cosine_recall@5
       value: 0.59
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.7
+      value: 0.74
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.5454496947645364
+      value: 0.5531504219817556
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.5129047619047619
+      value: 0.5029365079365079
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.5024388849786283
+      value: 0.5004093728081938
       name: Cosine Map@100
   - task:
       type: nano-beir
@@ -235,63 +235,63 @@ model-index:
       type: NanoBEIR_mean
     metrics:
     - type: cosine_accuracy@1
-      value: 0.44
+      value: 0.37
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.6000000000000001
+      value: 0.56
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.69
+      value: 0.63
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.79
+      value: 0.76
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.44
+      value: 0.37
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.2033333333333333
+      value: 0.19333333333333336
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.14
+      value: 0.132
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.081
+      value: 0.079
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.42500000000000004
+      value: 0.355
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.585
+      value: 0.555
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.675
+      value: 0.625
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.77
+      value: 0.75
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.5934854532996319
+      value: 0.5483162954426368
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.545797619047619
+      value: 0.4884920634920634
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.5442477549341866
+      value: 0.49198446733744927
       name: Cosine Map@100
 ---
 
-# SentenceTransformer based on Alibaba-NLP/gte-modernbert-base
+# SentenceTransformer based on thenlper/gte-small
 
-This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [Alibaba-NLP/gte-modernbert-base](https://huggingface.co/Alibaba-NLP/gte-modernbert-base). It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
+This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [thenlper/gte-small](https://huggingface.co/thenlper/gte-small). It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
 
 ## Model Details
 
 ### Model Description
 - **Model Type:** Sentence Transformer
-- **Base model:** [Alibaba-NLP/gte-modernbert-base](https://huggingface.co/Alibaba-NLP/gte-modernbert-base) <!-- at revision e7f32e3c00f91d699e8c43b53106206bcc72bb22 -->
+- **Base model:** [thenlper/gte-small](https://huggingface.co/thenlper/gte-small) <!-- at revision 17e1f347d17fe144873b1201da91788898c639cd -->
 - **Maximum Sequence Length:** 128 tokens
-- **Output Dimensionality:** 768 dimensions
+- **Output Dimensionality:** 384 dimensions
 - **Similarity Function:** Cosine Similarity
 <!-- - **Training Dataset:** Unknown -->
 <!-- - **Language:** Unknown -->
@@ -307,8 +307,9 @@ This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [A
 
 ```
 SentenceTransformer(
-  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False, 'architecture': 'ModernBertModel'})
-  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
+  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False, 'architecture': 'BertModel'})
+  (1): Pooling({'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})
+  (2): Normalize()
 )
 ```
 
@@ -336,14 +337,14 @@ sentences = [
 ]
 embeddings = model.encode(sentences)
 print(embeddings.shape)
-# [3, 768]
+# [3, 384]
 
 # Get the similarity scores for the embeddings
 similarities = model.similarity(embeddings, embeddings)
 print(similarities)
-# tensor([[1.0000, 0.4639, 0.5512],
-#         [0.4639, 1.0000, 0.1977],
-#         [0.5512, 0.1977, 1.0000]])
+# tensor([[0.9999, 0.7811, 0.4781],
+#         [0.7811, 1.0000, 0.3146],
+#         [0.4781, 0.3146, 1.0000]])
 ```
 
 <!--
@@ -381,21 +382,21 @@ You can finetune this model on your own dataset.
 
 | Metric              | NanoMSMARCO | NanoNQ     |
 |:--------------------|:------------|:-----------|
-| cosine_accuracy@1   | 0.46        | 0.42       |
-| cosine_accuracy@3   | 0.64        | 0.56       |
-| cosine_accuracy@5   | 0.76        | 0.62       |
-| cosine_accuracy@10  | 0.84        | 0.74       |
-| cosine_precision@1  | 0.46        | 0.42       |
-| cosine_precision@3  | 0.2133      | 0.1933     |
-| cosine_precision@5  | 0.152       | 0.128      |
-| cosine_precision@10 | 0.084       | 0.078      |
-| cosine_recall@1     | 0.46        | 0.39       |
-| cosine_recall@3     | 0.64        | 0.53       |
-| cosine_recall@5     | 0.76        | 0.59       |
-| cosine_recall@10    | 0.84        | 0.7        |
-| **cosine_ndcg@10**  | **0.6415**  | **0.5454** |
-| cosine_mrr@10       | 0.5787      | 0.5129     |
-| cosine_map@100      | 0.5861      | 0.5024     |
+| cosine_accuracy@1   | 0.32        | 0.42       |
+| cosine_accuracy@3   | 0.56        | 0.56       |
+| cosine_accuracy@5   | 0.66        | 0.6        |
+| cosine_accuracy@10  | 0.76        | 0.76       |
+| cosine_precision@1  | 0.32        | 0.42       |
+| cosine_precision@3  | 0.1867      | 0.2        |
+| cosine_precision@5  | 0.132       | 0.132      |
+| cosine_precision@10 | 0.076       | 0.082      |
+| cosine_recall@1     | 0.32        | 0.39       |
+| cosine_recall@3     | 0.56        | 0.55       |
+| cosine_recall@5     | 0.66        | 0.59       |
+| cosine_recall@10    | 0.76        | 0.74       |
+| **cosine_ndcg@10**  | **0.5435**  | **0.5532** |
+| cosine_mrr@10       | 0.474       | 0.5029     |
+| cosine_map@100      | 0.4836      | 0.5004     |
 
 #### Nano BEIR
 
@@ -413,21 +414,21 @@ You can finetune this model on your own dataset.
 
 | Metric              | Value      |
 |:--------------------|:-----------|
-| cosine_accuracy@1   | 0.44       |
-| cosine_accuracy@3   | 0.6        |
-| cosine_accuracy@5   | 0.69       |
-| cosine_accuracy@10  | 0.79       |
-| cosine_precision@1  | 0.44       |
-| cosine_precision@3  | 0.2033     |
-| cosine_precision@5  | 0.14       |
-| cosine_precision@10 | 0.081      |
-| cosine_recall@1     | 0.425      |
-| cosine_recall@3     | 0.585      |
-| cosine_recall@5     | 0.675      |
-| cosine_recall@10    | 0.77       |
-| **cosine_ndcg@10**  | **0.5935** |
-| cosine_mrr@10       | 0.5458     |
-| cosine_map@100      | 0.5442     |
+| cosine_accuracy@1   | 0.37       |
+| cosine_accuracy@3   | 0.56       |
+| cosine_accuracy@5   | 0.63       |
+| cosine_accuracy@10  | 0.76       |
+| cosine_precision@1  | 0.37       |
+| cosine_precision@3  | 0.1933     |
+| cosine_precision@5  | 0.132      |
+| cosine_precision@10 | 0.079      |
+| cosine_recall@1     | 0.355      |
+| cosine_recall@3     | 0.555      |
+| cosine_recall@5     | 0.625      |
+| cosine_recall@10    | 0.75       |
+| **cosine_ndcg@10**  | **0.5483** |
+| cosine_mrr@10       | 0.4885     |
+| cosine_map@100      | 0.492      |
 
 <!--
 ## Bias, Risks and Limitations
@@ -450,10 +451,10 @@ You can finetune this model on your own dataset.
 * Size: 90,000 training samples
 * Columns: <code>anchor</code>, <code>positive</code>, and <code>negative</code>
 * Approximate statistics based on the first 1000 samples:
-  |         | anchor                                                                          | positive                                                                            | negative                                                                            |
-  |:--------|:--------------------------------------------------------------------------------|:------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------|
-  | type    | string                                                                          | string                                                                              | string                                                                              |
-  | details | <ul><li>min: 4 tokens</li><li>mean: 9.3 tokens</li><li>max: 27 tokens</li></ul> | <ul><li>min: 20 tokens</li><li>mean: 79.39 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 17 tokens</li><li>mean: 78.77 tokens</li><li>max: 128 tokens</li></ul> |
+  |         | anchor                                                                           | positive                                                                            | negative                                                                            |
+  |:--------|:---------------------------------------------------------------------------------|:------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------|
+  | type    | string                                                                           | string                                                                              | string                                                                              |
+  | details | <ul><li>min: 4 tokens</li><li>mean: 9.18 tokens</li><li>max: 43 tokens</li></ul> | <ul><li>min: 19 tokens</li><li>mean: 78.75 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 17 tokens</li><li>mean: 77.97 tokens</li><li>max: 128 tokens</li></ul> |
 * Samples:
   | anchor                                            | positive                                                                                                                                                                                                                                                                                                                                                                          | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                |
   |:--------------------------------------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -476,10 +477,10 @@ You can finetune this model on your own dataset.
 * Size: 10,000 evaluation samples
 * Columns: <code>anchor</code>, <code>positive</code>, and <code>negative</code>
 * Approximate statistics based on the first 1000 samples:
-  |         | anchor                                                                           | positive                                                                           | negative                                                                            |
-  |:--------|:---------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:------------------------------------------------------------------------------------|
-  | type    | string                                                                           | string                                                                             | string                                                                              |
-  | details | <ul><li>min: 4 tokens</li><li>mean: 9.33 tokens</li><li>max: 21 tokens</li></ul> | <ul><li>min: 17 tokens</li><li>mean: 79.7 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 18 tokens</li><li>mean: 77.72 tokens</li><li>max: 128 tokens</li></ul> |
+  |         | anchor                                                                           | positive                                                                            | negative                                                                            |
+  |:--------|:---------------------------------------------------------------------------------|:------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------|
+  | type    | string                                                                           | string                                                                              | string                                                                              |
+  | details | <ul><li>min: 4 tokens</li><li>mean: 9.18 tokens</li><li>max: 25 tokens</li></ul> | <ul><li>min: 16 tokens</li><li>mean: 78.83 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 18 tokens</li><li>mean: 76.86 tokens</li><li>max: 128 tokens</li></ul> |
 * Samples:
   | anchor                                              | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      |
   |:----------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -501,7 +502,7 @@ You can finetune this model on your own dataset.
 - `eval_strategy`: steps
 - `per_device_train_batch_size`: 128
 - `per_device_eval_batch_size`: 128
-- `learning_rate`: 4e-05
+- `learning_rate`: 8e-05
 - `weight_decay`: 0.005
 - `max_steps`: 500
 - `warmup_ratio`: 0.1
@@ -530,7 +531,7 @@ You can finetune this model on your own dataset.
 - `gradient_accumulation_steps`: 1
 - `eval_accumulation_steps`: None
 - `torch_empty_cache_steps`: None
-- `learning_rate`: 4e-05
+- `learning_rate`: 8e-05
 - `weight_decay`: 0.005
 - `adam_beta1`: 0.9
 - `adam_beta2`: 0.999
@@ -644,9 +645,9 @@ You can finetune this model on your own dataset.
 ### Training Logs
 | Epoch  | Step | Training Loss | Validation Loss | NanoMSMARCO_cosine_ndcg@10 | NanoNQ_cosine_ndcg@10 | NanoBEIR_mean_cosine_ndcg@10 |
 |:------:|:----:|:-------------:|:---------------:|:--------------------------:|:---------------------:|:----------------------------:|
-| 0      | 0    | -             | 2.7123          | 0.6530                     | 0.6552                | 0.6541                       |
-| 0.3556 | 250  | 1.0557        | 0.8814          | 0.6765                     | 0.5338                | 0.6052                       |
-| 0.7112 | 500  | 0.8748        | 0.8640          | 0.6415                     | 0.5454                | 0.5935                       |
+| 0      | 0    | -             | 4.1735          | 0.6259                     | 0.6583                | 0.6421                       |
+| 0.3556 | 250  | 1.4992        | 0.9230          | 0.5606                     | 0.5505                | 0.5556                       |
+| 0.7112 | 500  | 1.047         | 0.9012          | 0.5435                     | 0.5532                | 0.5483                       |
 
 
 ### Framework Versions

config_sentence_transformers.json

@@ -1,4 +1,5 @@
 {
+  "model_type": "SentenceTransformer",
   "__version__": {
     "sentence_transformers": "5.2.0",
     "transformers": "4.57.3",
@@ -9,6 +10,5 @@
     "document": ""
   },
   "default_prompt_name": null,
-  "similarity_fn_name": "cosine",
-  "model_type": "SentenceTransformer"
+  "similarity_fn_name": "cosine"
 }

modules.json

@@ -10,5 +10,11 @@
     "name": "1",
     "path": "1_Pooling",
     "type": "sentence_transformers.models.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.models.Normalize"
   }
 ]