Add new SentenceTransformer model

1_Pooling/config.json

@@ -1,7 +1,7 @@
 {
-    "word_embedding_dimension": 384,
-    "pooling_mode_cls_token": false,
-    "pooling_mode_mean_tokens": true,
+    "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,

README.md

@@ -7,7 +7,7 @@ tags:
 - generated_from_trainer
 - dataset_size:111470
 - loss:MultipleNegativesRankingLoss
-base_model: sentence-transformers/all-MiniLM-L12-v2
+base_model: Alibaba-NLP/gte-modernbert-base
 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 sentence-transformers/all-MiniLM-L12-v2
+- name: SentenceTransformer based on Alibaba-NLP/gte-modernbert-base
   results:
   - task:
       type: information-retrieval
@@ -142,49 +142,49 @@ model-index:
       type: NanoMSMARCO
     metrics:
     - type: cosine_accuracy@1
-      value: 0.34
+      value: 0.38
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.5
+      value: 0.64
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.66
+      value: 0.76
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.78
+      value: 0.84
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.34
+      value: 0.38
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.16666666666666663
+      value: 0.21333333333333335
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.132
+      value: 0.15200000000000002
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.078
+      value: 0.08399999999999999
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.34
+      value: 0.38
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.5
+      value: 0.64
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.66
+      value: 0.76
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.78
+      value: 0.84
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.5446770528863051
+      value: 0.6071739753451822
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.4708571428571428
+      value: 0.5318571428571428
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.47884258431632043
+      value: 0.5383857612227555
       name: Cosine Map@100
   - task:
       type: information-retrieval
@@ -194,49 +194,49 @@ model-index:
       type: NanoNQ
     metrics:
     - type: cosine_accuracy@1
-      value: 0.5
+      value: 0.58
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.66
+      value: 0.74
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.7
+      value: 0.8
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.78
+      value: 0.84
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.5
+      value: 0.58
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.22666666666666668
+      value: 0.25333333333333335
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.14400000000000002
+      value: 0.16399999999999998
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.08199999999999999
+      value: 0.092
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.48
+      value: 0.55
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.64
+      value: 0.7
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.67
+      value: 0.74
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.74
+      value: 0.81
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.6136402968638738
+      value: 0.6914852313456867
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.5821666666666667
+      value: 0.6720555555555556
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.5768526974820034
+      value: 0.6484054934853453
       name: Cosine Map@100
   - task:
       type: nano-beir
@@ -246,63 +246,63 @@ model-index:
       type: NanoBEIR_mean
     metrics:
     - type: cosine_accuracy@1
-      value: 0.42000000000000004
+      value: 0.48
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.5800000000000001
+      value: 0.69
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.6799999999999999
+      value: 0.78
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.78
+      value: 0.84
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.42000000000000004
+      value: 0.48
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.19666666666666666
+      value: 0.23333333333333334
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.138
+      value: 0.158
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.07999999999999999
+      value: 0.088
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.41000000000000003
+      value: 0.465
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.5700000000000001
+      value: 0.6699999999999999
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.665
+      value: 0.75
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.76
+      value: 0.825
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.5791586748750894
+      value: 0.6493296033454345
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.5265119047619048
+      value: 0.6019563492063492
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.5278476408991619
+      value: 0.5933956273540504
       name: Cosine Map@100
 ---
 
-# SentenceTransformer based on sentence-transformers/all-MiniLM-L12-v2
+# SentenceTransformer based on Alibaba-NLP/gte-modernbert-base
 
-This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [sentence-transformers/all-MiniLM-L12-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L12-v2). 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.
+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.
 
 ## Model Details
 
 ### Model Description
 - **Model Type:** Sentence Transformer
-- **Base model:** [sentence-transformers/all-MiniLM-L12-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L12-v2) <!-- at revision 936af83a2ecce5fe87a09109ff5cbcefe073173a -->
+- **Base model:** [Alibaba-NLP/gte-modernbert-base](https://huggingface.co/Alibaba-NLP/gte-modernbert-base) <!-- at revision e7f32e3c00f91d699e8c43b53106206bcc72bb22 -->
 - **Maximum Sequence Length:** 128 tokens
-- **Output Dimensionality:** 384 dimensions
+- **Output Dimensionality:** 768 dimensions
 - **Similarity Function:** Cosine Similarity
 <!-- - **Training Dataset:** Unknown -->
 <!-- - **Language:** Unknown -->
@@ -318,9 +318,8 @@ This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [s
 
 ```
 SentenceTransformer(
-  (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()
+  (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})
 )
 ```
 
@@ -348,14 +347,14 @@ sentences = [
 ]
 embeddings = model.encode(sentences)
 print(embeddings.shape)
-# [3, 384]
+# [3, 768]
 
 # Get the similarity scores for the embeddings
 similarities = model.similarity(embeddings, embeddings)
 print(similarities)
-# tensor([[1.0000, 1.0000, 0.8845],
-#         [1.0000, 1.0000, 0.8845],
-#         [0.8845, 0.8845, 1.0000]])
+# tensor([[1.0000, 1.0000, 0.3030],
+#         [1.0000, 1.0000, 0.3030],
+#         [0.3030, 0.3030, 1.0001]])
 ```
 
 <!--
@@ -393,21 +392,21 @@ You can finetune this model on your own dataset.
 
 | Metric              | NanoMSMARCO | NanoNQ     |
 |:--------------------|:------------|:-----------|
-| cosine_accuracy@1   | 0.34        | 0.5        |
-| cosine_accuracy@3   | 0.5         | 0.66       |
-| cosine_accuracy@5   | 0.66        | 0.7        |
-| cosine_accuracy@10  | 0.78        | 0.78       |
-| cosine_precision@1  | 0.34        | 0.5        |
-| cosine_precision@3  | 0.1667      | 0.2267     |
-| cosine_precision@5  | 0.132       | 0.144      |
-| cosine_precision@10 | 0.078       | 0.082      |
-| cosine_recall@1     | 0.34        | 0.48       |
-| cosine_recall@3     | 0.5         | 0.64       |
-| cosine_recall@5     | 0.66        | 0.67       |
-| cosine_recall@10    | 0.78        | 0.74       |
-| **cosine_ndcg@10**  | **0.5447**  | **0.6136** |
-| cosine_mrr@10       | 0.4709      | 0.5822     |
-| cosine_map@100      | 0.4788      | 0.5769     |
+| cosine_accuracy@1   | 0.38        | 0.58       |
+| cosine_accuracy@3   | 0.64        | 0.74       |
+| cosine_accuracy@5   | 0.76        | 0.8        |
+| cosine_accuracy@10  | 0.84        | 0.84       |
+| cosine_precision@1  | 0.38        | 0.58       |
+| cosine_precision@3  | 0.2133      | 0.2533     |
+| cosine_precision@5  | 0.152       | 0.164      |
+| cosine_precision@10 | 0.084       | 0.092      |
+| cosine_recall@1     | 0.38        | 0.55       |
+| cosine_recall@3     | 0.64        | 0.7        |
+| cosine_recall@5     | 0.76        | 0.74       |
+| cosine_recall@10    | 0.84        | 0.81       |
+| **cosine_ndcg@10**  | **0.6072**  | **0.6915** |
+| cosine_mrr@10       | 0.5319      | 0.6721     |
+| cosine_map@100      | 0.5384      | 0.6484     |
 
 #### Nano BEIR
 
@@ -425,21 +424,21 @@ You can finetune this model on your own dataset.
 
 | Metric              | Value      |
 |:--------------------|:-----------|
-| cosine_accuracy@1   | 0.42       |
-| cosine_accuracy@3   | 0.58       |
-| cosine_accuracy@5   | 0.68       |
-| cosine_accuracy@10  | 0.78       |
-| cosine_precision@1  | 0.42       |
-| cosine_precision@3  | 0.1967     |
-| cosine_precision@5  | 0.138      |
-| cosine_precision@10 | 0.08       |
-| cosine_recall@1     | 0.41       |
-| cosine_recall@3     | 0.57       |
-| cosine_recall@5     | 0.665      |
-| cosine_recall@10    | 0.76       |
-| **cosine_ndcg@10**  | **0.5792** |
-| cosine_mrr@10       | 0.5265     |
-| cosine_map@100      | 0.5278     |
+| cosine_accuracy@1   | 0.48       |
+| cosine_accuracy@3   | 0.69       |
+| cosine_accuracy@5   | 0.78       |
+| cosine_accuracy@10  | 0.84       |
+| cosine_precision@1  | 0.48       |
+| cosine_precision@3  | 0.2333     |
+| cosine_precision@5  | 0.158      |
+| cosine_precision@10 | 0.088      |
+| cosine_recall@1     | 0.465      |
+| cosine_recall@3     | 0.67       |
+| cosine_recall@5     | 0.75       |
+| cosine_recall@10    | 0.825      |
+| **cosine_ndcg@10**  | **0.6493** |
+| cosine_mrr@10       | 0.602      |
+| cosine_map@100      | 0.5934     |
 
 <!--
 ## Bias, Risks and Limitations
@@ -465,7 +464,7 @@ 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: 13.22 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> |
+  | 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> |
 * Samples:
   | anchor                                                                                                  | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            | negative                                                                                                                                                                                                                                                                                                                                                                                                                               |
   |:--------------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -491,7 +490,7 @@ 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: 13.03 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> |
+  | 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> |
 * Samples:
   | anchor                                                                                                                                  | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        |
   |:----------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -513,9 +512,9 @@ 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`: 8e-05
-- `weight_decay`: 0.005
-- `max_steps`: 1125
+- `learning_rate`: 4e-05
+- `weight_decay`: 0.01
+- `max_steps`: 500
 - `warmup_ratio`: 0.1
 - `fp16`: True
 - `dataloader_drop_last`: True
@@ -542,14 +541,14 @@ You can finetune this model on your own dataset.
 - `gradient_accumulation_steps`: 1
 - `eval_accumulation_steps`: None
 - `torch_empty_cache_steps`: None
-- `learning_rate`: 8e-05
-- `weight_decay`: 0.005
+- `learning_rate`: 4e-05
+- `weight_decay`: 0.01
 - `adam_beta1`: 0.9
 - `adam_beta2`: 0.999
 - `adam_epsilon`: 1e-08
 - `max_grad_norm`: 1.0
 - `num_train_epochs`: 3.0
-- `max_steps`: 1125
+- `max_steps`: 500
 - `lr_scheduler_type`: linear
 - `lr_scheduler_kwargs`: {}
 - `warmup_ratio`: 0.1
@@ -654,14 +653,13 @@ You can finetune this model on your own dataset.
 </details>
 
 ### Training Logs
-| Epoch  | Step | Training Loss | Validation Loss | NanoMSMARCO_cosine_ndcg@10 | NanoNQ_cosine_ndcg@10 | NanoBEIR_mean_cosine_ndcg@10 |
-|:------:|:----:|:-------------:|:---------------:|:--------------------------:|:---------------------:|:----------------------------:|
-| 0      | 0    | -             | 0.1203          | 0.5887                     | 0.5786                | 0.5836                       |
-| 0.2874 | 250  | 0.094         | 0.0631          | 0.5536                     | 0.5611                | 0.5574                       |
-| 0.5747 | 500  | 0.0766        | 0.0586          | 0.5317                     | 0.5724                | 0.5521                       |
-| 0.8621 | 750  | 0.0674        | 0.0494          | 0.5357                     | 0.5675                | 0.5516                       |
-| 1.1494 | 1000 | 0.0491        | 0.0468          | 0.5447                     | 0.6136                | 0.5792                       |
+| Epoch      | Step    | Training Loss | Validation Loss | NanoMSMARCO_cosine_ndcg@10 | NanoNQ_cosine_ndcg@10 | NanoBEIR_mean_cosine_ndcg@10 |
+|:----------:|:-------:|:-------------:|:---------------:|:--------------------------:|:---------------------:|:----------------------------:|
+| 0          | 0       | -             | 0.4148          | 0.6530                     | 0.6552                | 0.6541                       |
+| 0.2874     | 250     | 0.0742        | 0.0408          | 0.6103                     | 0.6564                | 0.6333                       |
+| **0.5747** | **500** | **0.0398**    | **0.0376**      | **0.6072**                 | **0.6915**            | **0.6493**                   |
 
+* The bold row denotes the saved checkpoint.
 
 ### Framework Versions
 - Python: 3.10.18

config_sentence_transformers.json

@@ -4,11 +4,11 @@
     "transformers": "4.57.3",
     "pytorch": "2.9.1+cu128"
   },
-  "model_type": "SentenceTransformer",
   "prompts": {
     "query": "",
     "document": ""
   },
   "default_prompt_name": null,
-  "similarity_fn_name": "cosine"
+  "similarity_fn_name": "cosine",
+  "model_type": "SentenceTransformer"
 }

modules.json

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