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:90000
 - loss:MultipleNegativesRankingLoss
-base_model: sentence-transformers/all-MiniLM-L12-v2
+base_model: Alibaba-NLP/gte-modernbert-base
 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 sentence-transformers/all-MiniLM-L12-v2
+- name: SentenceTransformer based on Alibaba-NLP/gte-modernbert-base
   results:
   - task:
       type: information-retrieval
@@ -131,49 +131,49 @@ model-index:
       type: NanoMSMARCO
     metrics:
     - type: cosine_accuracy@1
-      value: 0.34
+      value: 0.46
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.56
+      value: 0.64
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.64
+      value: 0.76
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.74
+      value: 0.84
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.34
+      value: 0.46
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.18666666666666668
+      value: 0.21333333333333332
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.128
+      value: 0.15200000000000002
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.07400000000000001
+      value: 0.08399999999999999
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.34
+      value: 0.46
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.56
+      value: 0.64
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.64
+      value: 0.76
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.74
+      value: 0.84
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.5374407455925777
+      value: 0.6421089892290694
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.47276190476190466
+      value: 0.5794126984126984
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.48561801675925215
+      value: 0.586773135917494
       name: Cosine Map@100
   - task:
       type: information-retrieval
@@ -186,46 +186,46 @@ model-index:
       value: 0.4
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.5
+      value: 0.56
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.56
+      value: 0.62
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.7
+      value: 0.74
       name: Cosine Accuracy@10
     - type: cosine_precision@1
       value: 0.4
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.1733333333333333
+      value: 0.19333333333333333
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.12
+      value: 0.128
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.07400000000000001
+      value: 0.07800000000000001
       name: Cosine Precision@10
     - type: cosine_recall@1
       value: 0.37
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.47
+      value: 0.53
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.54
+      value: 0.59
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.66
+      value: 0.7
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.5075548380780193
+      value: 0.5380682898359656
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.47743650793650794
+      value: 0.502904761904762
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.46217171679843827
+      value: 0.4922434453305571
       name: Cosine Map@100
   - task:
       type: nano-beir
@@ -235,63 +235,63 @@ model-index:
       type: NanoBEIR_mean
     metrics:
     - type: cosine_accuracy@1
-      value: 0.37
+      value: 0.43000000000000005
       name: Cosine Accuracy@1
     - type: cosine_accuracy@3
-      value: 0.53
+      value: 0.6000000000000001
       name: Cosine Accuracy@3
     - type: cosine_accuracy@5
-      value: 0.6000000000000001
+      value: 0.69
       name: Cosine Accuracy@5
     - type: cosine_accuracy@10
-      value: 0.72
+      value: 0.79
       name: Cosine Accuracy@10
     - type: cosine_precision@1
-      value: 0.37
+      value: 0.43000000000000005
       name: Cosine Precision@1
     - type: cosine_precision@3
-      value: 0.18
+      value: 0.2033333333333333
       name: Cosine Precision@3
     - type: cosine_precision@5
-      value: 0.124
+      value: 0.14
       name: Cosine Precision@5
     - type: cosine_precision@10
-      value: 0.07400000000000001
+      value: 0.081
       name: Cosine Precision@10
     - type: cosine_recall@1
-      value: 0.355
+      value: 0.41500000000000004
       name: Cosine Recall@1
     - type: cosine_recall@3
-      value: 0.515
+      value: 0.585
       name: Cosine Recall@3
     - type: cosine_recall@5
-      value: 0.5900000000000001
+      value: 0.675
       name: Cosine Recall@5
     - type: cosine_recall@10
-      value: 0.7
+      value: 0.77
       name: Cosine Recall@10
     - type: cosine_ndcg@10
-      value: 0.5224977918352984
+      value: 0.5900886395325176
       name: Cosine Ndcg@10
     - type: cosine_mrr@10
-      value: 0.4750992063492063
+      value: 0.5411587301587302
       name: Cosine Mrr@10
     - type: cosine_map@100
-      value: 0.4738948667788452
+      value: 0.5395082906240256
       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 -->
@@ -307,9 +307,8 @@ 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})
 )
 ```
 
@@ -337,14 +336,14 @@ 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, 0.5519, 0.4707],
-#         [0.5519, 1.0001, 0.0831],
-#         [0.4707, 0.0831, 1.0000]])
+# tensor([[1.0000, 0.4641, 0.5503],
+#         [0.4641, 1.0000, 0.1977],
+#         [0.5503, 0.1977, 1.0000]])
 ```
 
 <!--
@@ -382,21 +381,21 @@ You can finetune this model on your own dataset.
 
 | Metric              | NanoMSMARCO | NanoNQ     |
 |:--------------------|:------------|:-----------|
-| cosine_accuracy@1   | 0.34        | 0.4        |
-| cosine_accuracy@3   | 0.56        | 0.5        |
-| cosine_accuracy@5   | 0.64        | 0.56       |
-| cosine_accuracy@10  | 0.74        | 0.7        |
-| cosine_precision@1  | 0.34        | 0.4        |
-| cosine_precision@3  | 0.1867      | 0.1733     |
-| cosine_precision@5  | 0.128       | 0.12       |
-| cosine_precision@10 | 0.074       | 0.074      |
-| cosine_recall@1     | 0.34        | 0.37       |
-| cosine_recall@3     | 0.56        | 0.47       |
-| cosine_recall@5     | 0.64        | 0.54       |
-| cosine_recall@10    | 0.74        | 0.66       |
-| **cosine_ndcg@10**  | **0.5374**  | **0.5076** |
-| cosine_mrr@10       | 0.4728      | 0.4774     |
-| cosine_map@100      | 0.4856      | 0.4622     |
+| cosine_accuracy@1   | 0.46        | 0.4        |
+| 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        | 0.4        |
+| 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        | 0.37       |
+| cosine_recall@3     | 0.64        | 0.53       |
+| cosine_recall@5     | 0.76        | 0.59       |
+| cosine_recall@10    | 0.84        | 0.7        |
+| **cosine_ndcg@10**  | **0.6421**  | **0.5381** |
+| cosine_mrr@10       | 0.5794      | 0.5029     |
+| cosine_map@100      | 0.5868      | 0.4922     |
 
 #### Nano BEIR
 
@@ -414,21 +413,21 @@ You can finetune this model on your own dataset.
 
 | Metric              | Value      |
 |:--------------------|:-----------|
-| cosine_accuracy@1   | 0.37       |
-| cosine_accuracy@3   | 0.53       |
-| cosine_accuracy@5   | 0.6        |
-| cosine_accuracy@10  | 0.72       |
-| cosine_precision@1  | 0.37       |
-| cosine_precision@3  | 0.18       |
-| cosine_precision@5  | 0.124      |
-| cosine_precision@10 | 0.074      |
-| cosine_recall@1     | 0.355      |
-| cosine_recall@3     | 0.515      |
-| cosine_recall@5     | 0.59       |
-| cosine_recall@10    | 0.7        |
-| **cosine_ndcg@10**  | **0.5225** |
-| cosine_mrr@10       | 0.4751     |
-| cosine_map@100      | 0.4739     |
+| cosine_accuracy@1   | 0.43       |
+| cosine_accuracy@3   | 0.6        |
+| cosine_accuracy@5   | 0.69       |
+| cosine_accuracy@10  | 0.79       |
+| cosine_precision@1  | 0.43       |
+| cosine_precision@3  | 0.2033     |
+| cosine_precision@5  | 0.14       |
+| cosine_precision@10 | 0.081      |
+| cosine_recall@1     | 0.415      |
+| cosine_recall@3     | 0.585      |
+| cosine_recall@5     | 0.675      |
+| cosine_recall@10    | 0.77       |
+| **cosine_ndcg@10**  | **0.5901** |
+| cosine_mrr@10       | 0.5412     |
+| cosine_map@100      | 0.5395     |
 
 <!--
 ## Bias, Risks and Limitations
@@ -451,10 +450,10 @@ 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: 9.18 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> |
+  |         | 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> |
 * Samples:
   | anchor                                            | positive                                                                                                                                                                                                                                                                                                                                                                          | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                |
   |:--------------------------------------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -477,10 +476,10 @@ 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: 9.18 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> |
+  |         | 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> |
 * Samples:
   | anchor                                              | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      |
   |:----------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -502,9 +501,9 @@ 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
+- `learning_rate`: 4e-05
 - `weight_decay`: 0.005
-- `max_steps`: 1687
+- `max_steps`: 500
 - `warmup_ratio`: 0.1
 - `fp16`: True
 - `dataloader_drop_last`: True
@@ -531,14 +530,14 @@ 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
+- `learning_rate`: 4e-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`: 1687
+- `max_steps`: 500
 - `lr_scheduler_type`: linear
 - `lr_scheduler_kwargs`: {}
 - `warmup_ratio`: 0.1
@@ -645,13 +644,9 @@ 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    | -             | 1.2073          | 0.5887                     | 0.5786                | 0.5836                       |
-| 0.3556 | 250  | 1.0978        | 0.9124          | 0.4990                     | 0.5159                | 0.5075                       |
-| 0.7112 | 500  | 1.0307        | 0.8904          | 0.5523                     | 0.4790                | 0.5157                       |
-| 1.0669 | 750  | 0.993         | 0.8816          | 0.5518                     | 0.4577                | 0.5047                       |
-| 1.4225 | 1000 | 0.8908        | 0.8770          | 0.5529                     | 0.5130                | 0.5329                       |
-| 1.7781 | 1250 | 0.8857        | 0.8703          | 0.5288                     | 0.5212                | 0.5250                       |
-| 2.1337 | 1500 | 0.8526        | 0.8677          | 0.5374                     | 0.5076                | 0.5225                       |
+| 0      | 0    | -             | 2.7123          | 0.6530                     | 0.6552                | 0.6541                       |
+| 0.3556 | 250  | 1.0558        | 0.8810          | 0.6601                     | 0.5409                | 0.6005                       |
+| 0.7112 | 500  | 0.8745        | 0.8635          | 0.6421                     | 0.5381                | 0.5901                       |
 
 
 ### Framework Versions

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"
   }
 ]