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:111470
 - loss:MultipleNegativesRankingLoss
-base_model: Alibaba-NLP/gte-modernbert-base
+base_model: thenlper/gte-small
 widget:
 - source_sentence: when was the first elephant brought to america
   sentences:
@@ -132,7 +132,7 @@ 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
@@ -142,49 +142,49 @@ model-index:
       type: NanoMSMARCO
     metrics:
     - type: cosine_accuracy@1
-      value: 0.24
+      value: 0.16
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.52
+      value: 0.32
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.56
+      value: 0.4
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.64
+      value: 0.54
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.24
+      value: 0.16
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.1733333333333333
+      value: 0.10666666666666666
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.11200000000000002
+      value: 0.08
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.06400000000000002
+      value: 0.05400000000000001
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.24
+      value: 0.16
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.52
+      value: 0.32
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.56
+      value: 0.4
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.64
+      value: 0.54
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.44801117912488453
+      value: 0.32698862634234876
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.3859444444444445
+      value: 0.2620793650793651
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.39907679444975275
+      value: 0.2747949118190278
       name: Cosine Map@100
   - task:
       type: information-retrieval
@@ -194,49 +194,49 @@ model-index:
       type: NanoNQ
     metrics:
     - type: cosine_accuracy@1
-      value: 0.3
+      value: 0.22
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.56
+      value: 0.44
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.66
+      value: 0.52
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.78
+      value: 0.64
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.3
+      value: 0.22
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.18666666666666668
+      value: 0.14666666666666664
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.132
+      value: 0.10400000000000001
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.08199999999999999
+      value: 0.06400000000000002
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.3
+      value: 0.22
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.56
+      value: 0.42
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.64
+      value: 0.49
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.76
+      value: 0.6
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.521342140364588
+      value: 0.39877036805974797
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.44460317460317456
+      value: 0.3438015873015873
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.4511292484432019
+      value: 0.3445409270682024
       name: Cosine Map@100
   - task:
       type: nano-beir
@@ -246,63 +246,63 @@ model-index:
       type: NanoBEIR_mean
     metrics:
     - type: cosine_accuracy@1
-      value: 0.27
+      value: 0.19
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.54
+      value: 0.38
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.6100000000000001
+      value: 0.46
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.71
+      value: 0.5900000000000001
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.27
+      value: 0.19
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.18
+      value: 0.12666666666666665
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.12200000000000001
+      value: 0.092
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.07300000000000001
+      value: 0.05900000000000001
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.27
+      value: 0.19
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.54
+      value: 0.37
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.6000000000000001
+      value: 0.445
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.7
+      value: 0.5700000000000001
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.4846766597447363
+      value: 0.3628794972010484
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.41527380952380955
+      value: 0.30294047619047615
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.4251030214464773
+      value: 0.3096679194436151
       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 -->
@@ -318,8 +318,9 @@ 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()
 )
 ```
 
@@ -347,14 +348,14 @@ 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([[1.0001, 1.0001, 0.1242],
-#         [1.0001, 1.0001, 0.1242],
-#         [0.1242, 0.1242, 1.0001]])
+# tensor([[1.0001, 1.0001, 0.4386],
+#         [1.0001, 1.0001, 0.4386],
+#         [0.4386, 0.4386, 1.0000]])
 ```
 
 <!--
@@ -392,21 +393,21 @@ You can finetune this model on your own dataset.
 
 | Metric              | NanoMSMARCO | NanoNQ     |
 |:--------------------|:------------|:-----------|
-| cosine_accuracy@1   | 0.24        | 0.3        |
-| cosine_accuracy@3   | 0.52        | 0.56       |
-| cosine_accuracy@5   | 0.56        | 0.66       |
-| cosine_accuracy@10  | 0.64        | 0.78       |
-| cosine_precision@1  | 0.24        | 0.3        |
-| cosine_precision@3  | 0.1733      | 0.1867     |
-| cosine_precision@5  | 0.112       | 0.132      |
-| cosine_precision@10 | 0.064       | 0.082      |
-| cosine_recall@1     | 0.24        | 0.3        |
-| cosine_recall@3     | 0.52        | 0.56       |
-| cosine_recall@5     | 0.56        | 0.64       |
-| cosine_recall@10    | 0.64        | 0.76       |
-| **cosine_ndcg@10**  | **0.448**   | **0.5213** |
-| cosine_mrr@10       | 0.3859      | 0.4446     |
-| cosine_map@100      | 0.3991      | 0.4511     |
+| cosine_accuracy@1   | 0.16        | 0.22       |
+| cosine_accuracy@3   | 0.32        | 0.44       |
+| cosine_accuracy@5   | 0.4         | 0.52       |
+| cosine_accuracy@10  | 0.54        | 0.64       |
+| cosine_precision@1  | 0.16        | 0.22       |
+| cosine_precision@3  | 0.1067      | 0.1467     |
+| cosine_precision@5  | 0.08        | 0.104      |
+| cosine_precision@10 | 0.054       | 0.064      |
+| cosine_recall@1     | 0.16        | 0.22       |
+| cosine_recall@3     | 0.32        | 0.42       |
+| cosine_recall@5     | 0.4         | 0.49       |
+| cosine_recall@10    | 0.54        | 0.6        |
+| **cosine_ndcg@10**  | **0.327**   | **0.3988** |
+| cosine_mrr@10       | 0.2621      | 0.3438     |
+| cosine_map@100      | 0.2748      | 0.3445     |
 
 #### Nano BEIR
 
@@ -424,21 +425,21 @@ You can finetune this model on your own dataset.
 
 | Metric              | Value      |
 |:--------------------|:-----------|
-| cosine_accuracy@1   | 0.27       |
-| cosine_accuracy@3   | 0.54       |
-| cosine_accuracy@5   | 0.61       |
-| cosine_accuracy@10  | 0.71       |
-| cosine_precision@1  | 0.27       |
-| cosine_precision@3  | 0.18       |
-| cosine_precision@5  | 0.122      |
-| cosine_precision@10 | 0.073      |
-| cosine_recall@1     | 0.27       |
-| cosine_recall@3     | 0.54       |
-| cosine_recall@5     | 0.6        |
-| cosine_recall@10    | 0.7        |
-| **cosine_ndcg@10**  | **0.4847** |
-| cosine_mrr@10       | 0.4153     |
-| cosine_map@100      | 0.4251     |
+| cosine_accuracy@1   | 0.19       |
+| cosine_accuracy@3   | 0.38       |
+| cosine_accuracy@5   | 0.46       |
+| cosine_accuracy@10  | 0.59       |
+| cosine_precision@1  | 0.19       |
+| cosine_precision@3  | 0.1267     |
+| cosine_precision@5  | 0.092      |
+| cosine_precision@10 | 0.059      |
+| cosine_recall@1     | 0.19       |
+| cosine_recall@3     | 0.37       |
+| cosine_recall@5     | 0.445      |
+| cosine_recall@10    | 0.57       |
+| **cosine_ndcg@10**  | **0.3629** |
+| cosine_mrr@10       | 0.3029     |
+| cosine_map@100      | 0.3097     |
 
 <!--
 ## Bias, Risks and Limitations
@@ -464,7 +465,7 @@ You can finetune this model on your own dataset.
   |         | anchor                                                                            | positive                                                                           | negative                                                                           |
   |:--------|:----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|
   | type    | string                                                                            | string                                                                             | string                                                                             |
-  | details | <ul><li>min: 6 tokens</li><li>mean: 13.83 tokens</li><li>max: 45 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 91.42 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 7 tokens</li><li>mean: 90.36 tokens</li><li>max: 128 tokens</li></ul> |
+  | details | <ul><li>min: 6 tokens</li><li>mean: 13.22 tokens</li><li>max: 44 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 90.67 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 7 tokens</li><li>mean: 89.65 tokens</li><li>max: 128 tokens</li></ul> |
 * Samples:
   | anchor                                                                                                  | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            | negative                                                                                                                                                                                                                                                                                                                                                                                                                               |
   |:--------------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -490,7 +491,7 @@ You can finetune this model on your own dataset.
   |         | anchor                                                                            | positive                                                                           | negative                                                                           |
   |:--------|:----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|
   | type    | string                                                                            | string                                                                             | string                                                                             |
-  | details | <ul><li>min: 6 tokens</li><li>mean: 13.69 tokens</li><li>max: 45 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 90.17 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 89.67 tokens</li><li>max: 128 tokens</li></ul> |
+  | details | <ul><li>min: 6 tokens</li><li>mean: 13.03 tokens</li><li>max: 44 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 89.36 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 4 tokens</li><li>mean: 88.87 tokens</li><li>max: 128 tokens</li></ul> |
 * Samples:
   | anchor                                                                                                                                  | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        |
   |:----------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -512,9 +513,9 @@ 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
-- `weight_decay`: 0.01
-- `max_steps`: 703
+- `learning_rate`: 8e-05
+- `weight_decay`: 0.005
+- `max_steps`: 3375
 - `warmup_ratio`: 0.1
 - `fp16`: True
 - `dataloader_drop_last`: True
@@ -541,14 +542,14 @@ 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
-- `weight_decay`: 0.01
+- `learning_rate`: 8e-05
+- `weight_decay`: 0.005
 - `adam_beta1`: 0.9
 - `adam_beta2`: 0.999
 - `adam_epsilon`: 1e-08
 - `max_grad_norm`: 1.0
 - `num_train_epochs`: 3.0
-- `max_steps`: 703
+- `max_steps`: 3375
 - `lr_scheduler_type`: linear
 - `lr_scheduler_kwargs`: {}
 - `warmup_ratio`: 0.1
@@ -655,9 +656,20 @@ 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    | -             | 4.3452          | 0.6530                     | 0.6552                | 0.6541                       |
-| 0.2874 | 250  | 3.1166        | 2.9191          | 0.4629                     | 0.5508                | 0.5069                       |
-| 0.5747 | 500  | 2.9043        | 2.8945          | 0.4480                     | 0.5213                | 0.4847                       |
+| 0      | 0    | -             | 4.9445          | 0.6259                     | 0.6583                | 0.6421                       |
+| 0.2874 | 250  | 3.6887        | 3.0013          | 0.4676                     | 0.4424                | 0.4550                       |
+| 0.5747 | 500  | 3.0661        | 2.9415          | 0.4647                     | 0.4688                | 0.4667                       |
+| 0.8621 | 750  | 3.0125        | 2.9161          | 0.3994                     | 0.4479                | 0.4237                       |
+| 1.1494 | 1000 | 2.9691        | 2.9044          | 0.3845                     | 0.4090                | 0.3968                       |
+| 1.4368 | 1250 | 2.9407        | 2.8981          | 0.3614                     | 0.3858                | 0.3736                       |
+| 1.7241 | 1500 | 2.9321        | 2.8893          | 0.3182                     | 0.3811                | 0.3496                       |
+| 2.0115 | 1750 | 2.9227        | 2.8817          | 0.3444                     | 0.3973                | 0.3708                       |
+| 2.2989 | 2000 | 2.8854        | 2.8807          | 0.3088                     | 0.3730                | 0.3409                       |
+| 2.5862 | 2250 | 2.8832        | 2.8744          | 0.3251                     | 0.3968                | 0.3610                       |
+| 2.8736 | 2500 | 2.8857        | 2.8730          | 0.3504                     | 0.4101                | 0.3802                       |
+| 3.1609 | 2750 | 2.8677        | 2.8714          | 0.3233                     | 0.4021                | 0.3627                       |
+| 3.4483 | 3000 | 2.86          | 2.8697          | 0.3239                     | 0.4106                | 0.3673                       |
+| 3.7356 | 3250 | 2.8584        | 2.8686          | 0.3270                     | 0.3988                | 0.3629                       |
 
 
 ### 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"
   }
 ]