Add new SentenceTransformer model

README.md

@@ -5,38 +5,108 @@ tags:
 - feature-extraction
 - dense
 - generated_from_trainer
-- dataset_size:100000
+- dataset_size:90000
 - loss:MultipleNegativesRankingLoss
 base_model: prajjwal1/bert-small
 widget:
-- source_sentence: How do I calculate IQ?
+- source_sentence: How do I cope with my depression to keep my girlfriend?
   sentences:
-  - What is the easiest way to know my IQ?
-  - How do I calculate not IQ ?
-  - What are some creative and innovative business ideas with less investment in India?
-- source_sentence: How can I learn martial arts in my home?
+  - How do you cope with depression?
+  - How do I cope with my depression to keep my girlfriend?
+  - What is the significance of Nevanlinna theory?
+- source_sentence: Which is the best college for MBA in Delhi?
   sentences:
-  - How can I learn martial arts by myself?
-  - What are the advantages and disadvantages of investing in gold?
-  - Can people see that I have looked at their pictures on instagram if I am not following
-    them?
-- source_sentence: When Enterprise picks you up do you have to take them back?
+  - Who played the character of 'Russ' in friends?
+  - What will be the picture quality if a Standard STB is Connected to a Full HD 40"
+    Led TV?
+  - Which is the best college to do an MBA in Delhi?
+- source_sentence: What is poison mailbox?
   sentences:
-  - Are there any software Training institute in Tuticorin?
-  - When Enterprise picks you up do you have to take them back?
-  - When Enterprise picks you up do them have to take youback?
-- source_sentence: What are some non-capital goods?
+  - What are the differences between a concentric and eccentric movement?
+  - What is poison mailbox?
+  - What is not poison mailbox ?
+- source_sentence: What are the examples of a presence of mind?
   sentences:
-  - What are capital goods?
-  - How is the value of [math]\pi[/math] calculated?
-  - What are some non-capital goods?
-- source_sentence: What is the QuickBooks technical support phone number in New York?
+  - What are not the examples of a presence of mind ?
+  - What's the point of dressing well?
+  - What are some of the greatest examples of the presence of mind?
+- source_sentence: What are some tips on making it through the job interview process
+    at Opus Bank?
   sentences:
-  - What caused the Great Depression?
-  - Can I apply for PR in Canada?
-  - Which is the best QuickBooks Hosting Support Number in New York?
+  - I've got an online coupon for Domino's pizza through the freecharge app. Is it
+    necessary to use that coupon only when I order online?
+  - What are not some tips on making it through the job interview process at Opus
+    Bank ?
+  - What are some tips on making it through the job interview process at Opus Bank?
 pipeline_tag: sentence-similarity
 library_name: sentence-transformers
+metrics:
+- cosine_accuracy@1
+- cosine_accuracy@3
+- cosine_accuracy@5
+- cosine_precision@1
+- cosine_precision@3
+- cosine_precision@5
+- cosine_recall@1
+- cosine_recall@3
+- cosine_recall@5
+- cosine_ndcg@10
+- cosine_mrr@1
+- cosine_mrr@5
+- cosine_mrr@10
+- cosine_map@100
+model-index:
+- name: SentenceTransformer based on prajjwal1/bert-small
+  results:
+  - task:
+      type: information-retrieval
+      name: Information Retrieval
+    dataset:
+      name: val
+      type: val
+    metrics:
+    - type: cosine_accuracy@1
+      value: 0.903
+      name: Cosine Accuracy@1
+    - type: cosine_accuracy@3
+      value: 0.9652
+      name: Cosine Accuracy@3
+    - type: cosine_accuracy@5
+      value: 0.9802
+      name: Cosine Accuracy@5
+    - type: cosine_precision@1
+      value: 0.903
+      name: Cosine Precision@1
+    - type: cosine_precision@3
+      value: 0.32173333333333337
+      name: Cosine Precision@3
+    - type: cosine_precision@5
+      value: 0.19603999999999996
+      name: Cosine Precision@5
+    - type: cosine_recall@1
+      value: 0.903
+      name: Cosine Recall@1
+    - type: cosine_recall@3
+      value: 0.9652
+      name: Cosine Recall@3
+    - type: cosine_recall@5
+      value: 0.9802
+      name: Cosine Recall@5
+    - type: cosine_ndcg@10
+      value: 0.9497950442756341
+      name: Cosine Ndcg@10
+    - type: cosine_mrr@1
+      value: 0.903
+      name: Cosine Mrr@1
+    - type: cosine_mrr@5
+      value: 0.93429
+      name: Cosine Mrr@5
+    - type: cosine_mrr@10
+      value: 0.93595873015873
+      name: Cosine Mrr@10
+    - type: cosine_map@100
+      value: 0.9364845314523799
+      name: Cosine Map@100
 ---
 
 # SentenceTransformer based on prajjwal1/bert-small
@@ -85,12 +155,12 @@ Then you can load this model and run inference.
 from sentence_transformers import SentenceTransformer
 
 # Download from the 🤗 Hub
-model = SentenceTransformer("sentence_transformers_model_id")
+model = SentenceTransformer("redis/model-b-structured")
 # Run inference
 sentences = [
-    'What is the QuickBooks technical support phone number in New York?',
-    'Which is the best QuickBooks Hosting Support Number in New York?',
-    'Can I apply for PR in Canada?',
+    'What are some tips on making it through the job interview process at Opus Bank?',
+    'What are some tips on making it through the job interview process at Opus Bank?',
+    'What are not some tips on making it through the job interview process at Opus Bank ?',
 ]
 embeddings = model.encode(sentences)
 print(embeddings.shape)
@@ -99,9 +169,9 @@ print(embeddings.shape)
 # Get the similarity scores for the embeddings
 similarities = model.similarity(embeddings, embeddings)
 print(similarities)
-# tensor([[1.0000, 0.8563, 0.0594],
-#         [0.8563, 1.0000, 0.1245],
-#         [0.0594, 0.1245, 1.0000]])
+# tensor([[1.0000, 1.0000, 0.1451],
+#         [1.0000, 1.0000, 0.1451],
+#         [0.1451, 0.1451, 1.0000]])
 ```
 
 <!--
@@ -128,6 +198,32 @@ You can finetune this model on your own dataset.
 *List how the model may foreseeably be misused and address what users ought not to do with the model.*
 -->
 
+## Evaluation
+
+### Metrics
+
+#### Information Retrieval
+
+* Dataset: `val`
+* Evaluated with [<code>InformationRetrievalEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator)
+
+| Metric             | Value      |
+|:-------------------|:-----------|
+| cosine_accuracy@1  | 0.903      |
+| cosine_accuracy@3  | 0.9652     |
+| cosine_accuracy@5  | 0.9802     |
+| cosine_precision@1 | 0.903      |
+| cosine_precision@3 | 0.3217     |
+| cosine_precision@5 | 0.196      |
+| cosine_recall@1    | 0.903      |
+| cosine_recall@3    | 0.9652     |
+| cosine_recall@5    | 0.9802     |
+| **cosine_ndcg@10** | **0.9498** |
+| cosine_mrr@1       | 0.903      |
+| cosine_mrr@5       | 0.9343     |
+| cosine_mrr@10      | 0.936      |
+| cosine_map@100     | 0.9365     |
+
 <!--
 ## Bias, Risks and Limitations
 
@@ -146,19 +242,45 @@ You can finetune this model on your own dataset.
 
 #### Unnamed Dataset
 
-* Size: 100,000 training samples
-* Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>sentence_2</code>
+* 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: 6 tokens</li><li>mean: 15.63 tokens</li><li>max: 75 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 15.77 tokens</li><li>max: 75 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 16.2 tokens</li><li>max: 75 tokens</li></ul> |
+* Samples:
+  | anchor                                                   | positive                                                 | negative                                                                    |
+  |:---------------------------------------------------------|:---------------------------------------------------------|:----------------------------------------------------------------------------|
+  | <code>How long did it take to develop Pokémon GO?</code> | <code>How long did it take to develop Pokémon GO?</code> | <code>Can I take more than one gym in Pokémon GO?</code>                    |
+  | <code>How bad is 6/18 eyesight?</code>                   | <code>How bad is 6/18 eyesight?</code>                   | <code>How was bad eyesight dealt with in ancient and medieval times?</code> |
+  | <code>How can I do learn speaking English easily?</code> | <code>How can I learn speaking English easily?</code>    | <code>How can English do learn speaking Ieasily?</code>                     |
+* Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
+  ```json
+  {
+      "scale": 20.0,
+      "similarity_fct": "cos_sim",
+      "gather_across_devices": false
+  }
+  ```
+
+### Evaluation Dataset
+
+#### Unnamed Dataset
+
+* Size: 5,000 evaluation samples
+* Columns: <code>anchor</code>, <code>positive</code>, and <code>negative</code>
 * Approximate statistics based on the first 1000 samples:
-  |         | sentence_0                                                                        | sentence_1                                                                        | sentence_2                                                                        |
+  |         | anchor                                                                            | positive                                                                          | negative                                                                          |
   |:--------|:----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|
   | type    | string                                                                            | string                                                                            | string                                                                            |
-  | details | <ul><li>min: 6 tokens</li><li>mean: 15.79 tokens</li><li>max: 66 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 15.68 tokens</li><li>max: 66 tokens</li></ul> | <ul><li>min: 7 tokens</li><li>mean: 16.37 tokens</li><li>max: 67 tokens</li></ul> |
+  | details | <ul><li>min: 6 tokens</li><li>mean: 15.65 tokens</li><li>max: 49 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 15.69 tokens</li><li>max: 49 tokens</li></ul> | <ul><li>min: 6 tokens</li><li>mean: 16.66 tokens</li><li>max: 55 tokens</li></ul> |
 * Samples:
-  | sentence_0                                                       | sentence_1                                                       | sentence_2                                                                        |
-  |:-----------------------------------------------------------------|:-----------------------------------------------------------------|:----------------------------------------------------------------------------------|
-  | <code>Is masturbating bad for boys?</code>                       | <code>Is masturbating bad for boys?</code>                       | <code>How harmful or unhealthy is masturbation?</code>                            |
-  | <code>Does a train engine move in reverse?</code>                | <code>Does a train engine move in reverse?</code>                | <code>Time moves forward, not in reverse. Doesn't that make time a vector?</code> |
-  | <code>What is the most badass thing anyone has ever done?</code> | <code>What is the most badass thing anyone has ever done?</code> | <code>anyone is the most badass thing Whathas ever done?</code>                   |
+  | anchor                                                                                                                                                             | positive                                                                                                                                                           | negative                                                                                                                                                               |
+  |:-------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+  | <code>What's it like working in IT for Goldman Sachs?</code>                                                                                                       | <code>What's it like working in IT for Goldman Sachs?</code>                                                                                                       | <code>it 's Whatlike working in IT for Goldman Sachs?</code>                                                                                                           |
+  | <code>Will time travel be possible in future?</code>                                                                                                               | <code>Is time travel still theorized as being possible?</code>                                                                                                     | <code>Will time travel be not possible in future ?</code>                                                                                                              |
+  | <code>For creating a software based service for SME’s, we need to tie up with a bank. Need the best way to contact the right person in big banks like HDFC.</code> | <code>For creating a software based service for SME’s, we need to tie up with a bank. Need the best way to contact the right person in big banks like HDFC.</code> | <code>For creating a software based service for SME ’s , we need to tie up with a bank . Need the right way to contact the best person in big banks like HDFC .</code> |
 * Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
   ```json
   {
@@ -171,36 +293,49 @@ You can finetune this model on your own dataset.
 ### Training Hyperparameters
 #### Non-Default Hyperparameters
 
-- `per_device_train_batch_size`: 64
-- `per_device_eval_batch_size`: 64
+- `eval_strategy`: steps
+- `per_device_train_batch_size`: 256
+- `per_device_eval_batch_size`: 256
+- `learning_rate`: 2e-05
+- `weight_decay`: 0.001
+- `max_steps`: 1053
+- `warmup_ratio`: 0.1
 - `fp16`: True
-- `multi_dataset_batch_sampler`: round_robin
+- `dataloader_drop_last`: True
+- `dataloader_num_workers`: 1
+- `dataloader_prefetch_factor`: 1
+- `load_best_model_at_end`: True
+- `optim`: adamw_torch
+- `ddp_find_unused_parameters`: False
+- `push_to_hub`: True
+- `hub_model_id`: redis/model-b-structured
+- `eval_on_start`: True
 
 #### All Hyperparameters
 <details><summary>Click to expand</summary>
 
 - `overwrite_output_dir`: False
 - `do_predict`: False
-- `eval_strategy`: no
+- `eval_strategy`: steps
 - `prediction_loss_only`: True
-- `per_device_train_batch_size`: 64
-- `per_device_eval_batch_size`: 64
+- `per_device_train_batch_size`: 256
+- `per_device_eval_batch_size`: 256
 - `per_gpu_train_batch_size`: None
 - `per_gpu_eval_batch_size`: None
 - `gradient_accumulation_steps`: 1
 - `eval_accumulation_steps`: None
 - `torch_empty_cache_steps`: None
-- `learning_rate`: 5e-05
-- `weight_decay`: 0.0
+- `learning_rate`: 2e-05
+- `weight_decay`: 0.001
 - `adam_beta1`: 0.9
 - `adam_beta2`: 0.999
 - `adam_epsilon`: 1e-08
-- `max_grad_norm`: 1
-- `num_train_epochs`: 3
-- `max_steps`: -1
+- `max_grad_norm`: 1.0
+- `num_train_epochs`: 3.0
+- `max_steps`: 1053
 - `lr_scheduler_type`: linear
 - `lr_scheduler_kwargs`: {}
-- `warmup_ratio`: 0.0
+- `warmup_ratio`: 0.1
 - `warmup_steps`: 0
 - `log_level`: passive
 - `log_level_replica`: warning
@@ -228,14 +363,14 @@ You can finetune this model on your own dataset.
 - `tpu_num_cores`: None
 - `tpu_metrics_debug`: False
 - `debug`: []
-- `dataloader_drop_last`: False
-- `dataloader_num_workers`: 0
-- `dataloader_prefetch_factor`: None
+- `dataloader_drop_last`: True
+- `dataloader_num_workers`: 1
+- `dataloader_prefetch_factor`: 1
 - `past_index`: -1
 - `disable_tqdm`: False
 - `remove_unused_columns`: True
 - `label_names`: None
-- `load_best_model_at_end`: False
+- `load_best_model_at_end`: True
 - `ignore_data_skip`: False
 - `fsdp`: []
 - `fsdp_min_num_params`: 0
@@ -245,23 +380,23 @@ You can finetune this model on your own dataset.
 - `parallelism_config`: None
 - `deepspeed`: None
 - `label_smoothing_factor`: 0.0
-- `optim`: adamw_torch_fused
+- `optim`: adamw_torch
 - `optim_args`: None
 - `adafactor`: False
 - `group_by_length`: False
 - `length_column_name`: length
 - `project`: huggingface
 - `trackio_space_id`: trackio
-- `ddp_find_unused_parameters`: None
+- `ddp_find_unused_parameters`: False
 - `ddp_bucket_cap_mb`: None
 - `ddp_broadcast_buffers`: False
 - `dataloader_pin_memory`: True
 - `dataloader_persistent_workers`: False
 - `skip_memory_metrics`: True
 - `use_legacy_prediction_loop`: False
-- `push_to_hub`: False
+- `push_to_hub`: True
 - `resume_from_checkpoint`: None
-- `hub_model_id`: None
+- `hub_model_id`: redis/model-b-structured
 - `hub_strategy`: every_save
 - `hub_private_repo`: None
 - `hub_always_push`: False
@@ -288,32 +423,35 @@ You can finetune this model on your own dataset.
 - `neftune_noise_alpha`: None
 - `optim_target_modules`: None
 - `batch_eval_metrics`: False
-- `eval_on_start`: False
+- `eval_on_start`: True
 - `use_liger_kernel`: False
 - `liger_kernel_config`: None
 - `eval_use_gather_object`: False
 - `average_tokens_across_devices`: True
 - `prompts`: None
 - `batch_sampler`: batch_sampler
-- `multi_dataset_batch_sampler`: round_robin
+- `multi_dataset_batch_sampler`: proportional
 - `router_mapping`: {}
 - `learning_rate_mapping`: {}
 
 </details>
 
 ### Training Logs
-| Epoch  | Step | Training Loss |
-|:------:|:----:|:-------------:|
-| 0.3199 | 500  | 0.4294        |
-| 0.6398 | 1000 | 0.1268        |
-| 0.9597 | 1500 | 0.1           |
-| 1.2796 | 2000 | 0.0792        |
-| 1.5995 | 2500 | 0.0706        |
-| 1.9194 | 3000 | 0.0687        |
-| 2.2393 | 3500 | 0.0584        |
-| 2.5592 | 4000 | 0.057         |
-| 2.8791 | 4500 | 0.0581        |
-
+| Epoch     | Step     | Training Loss | Validation Loss | val_cosine_ndcg@10 |
+|:---------:|:--------:|:-------------:|:---------------:|:------------------:|
+| 0         | 0        | -             | 1.7213          | 0.8444             |
+| 0.2849    | 100      | 1.0498        | -               | -                  |
+| 0.5698    | 200      | 0.2458        | -               | -                  |
+| 0.8547    | 300      | 0.1967        | -               | -                  |
+| 1.1396    | 400      | 0.1731        | -               | -                  |
+| 1.4245    | 500      | 0.158         | 0.1040          | 0.9472             |
+| 1.7094    | 600      | 0.1503        | -               | -                  |
+| 1.9943    | 700      | 0.14          | -               | -                  |
+| 2.2792    | 800      | 0.1375        | -               | -                  |
+| 2.5641    | 900      | 0.1301        | -               | -                  |
+| **2.849** | **1000** | **0.1234**    | **0.0925**      | **0.9498**         |
+
+* The bold row denotes the saved checkpoint.
 
 ### Framework Versions
 - Python: 3.10.18

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4cc9a88efd8d3822ebf670b0c7aef92d3a665b705509767b6b8654e668f60314
+oid sha256:9165bec07d721d490566906032416157feb6461c69983b2c810cd24685e50198
 size 114011616