Blame for v1.0/experiments/robustness/models/dm_s14/README.md

---
tags:
- sentence-transformers
- sentence-similarity
- feature-extraction
- dense
- generated_from_trainer
- dataset_size:59612
- loss:CachedMultipleNegativesRankingLoss
- loss:MultipleNegativesRankingLoss
base_model: intfloat/multilingual-e5-base
widget:
- source_sentence: 'query: Point of view regarding UK reform demands

    On 11 November 2015, the UK Prime Minister, David Cameron, formally set out his
    demands in a letter to the President of the European Council, stating that four
    objectives lie at the heart of the UK’s renegotiations. What is the Council’s
    perception in relation to the goals for reforming the UK’s membership of the EU,
    including the restrictions on benefits for people coming to the UK?'
  sentences:
  - 'passage: The Council dealt with the UK reform demands as part of the preparations
    for the European Council. The Council supported the aim of finding mutually satisfactory
    solutions in all four areas mentioned in the UK Prime Minister''s letter of 10
    November 2015. However, the Council is not the institution in charge of finding
    a solution to this issue. The New Settlement for the United Kingdom within the
    European Union was agreed at the level of Heads of State or Government and the
    President of the Commission at the meeting of the European Council of 18 and 19
    February 2016. The set of arrangements constituting this settlement was published
    in the Official Journal of the European Union[1].

    [1] OJ C 69 I, 23.2.2016, p. 1.'
  - 'passage: President Juncker has repeatedly said that he wants to achieve a deal
    that is fair for the UK and fair for all other Member States. The Commission and
    the Council are actively working on the requests of the UK as set out in Prime
    Minister Cameron''s letter to President Tusk of 10 November 2015.

    The Commission does not speculate on the results of the UK referendum.'
  - 'passage: The Commission is regularly informed about the situation at the Spanish-Gibraltar
    border, including on delays reported by citizens upon entry or exit to Spain.

    On 30 July 2014 the Commission addressed requests for additional measures to both
    Spain and the United Kingdom to be implemented without delay in order to reduce
    the waiting time. One of the important measures requested from Spain is to optimise
    the risk-based profiling and to adapt the checks accordingly. The Commission recently
    received the reply from Spain. It will proceed to its analysis before assessing
    the situation at this border again.'
- source_sentence: 'query: Compliance with Directive 2012/47/EU

    In response to Question [REF], Mr Nelli Feroci answered on behalf of the Commission:
    ''On the basis of the first analysis of transposition measures notified by Spain
    as well as of available information, the Commission considers prima facie that
    Directive 2012/47/EU has been correctly transposed by Spain.''

    1. Has the Commission now completed the full analysis of the transposition measures
    notified by Spain?

    2. Is the Commission satisfied with the result?'
  sentences:
  - 'passage: It is not for the Council to comment on a legislative act that has been
    adopted by the European Parliament and by the Council in the ordinary legislative
    procedure.'
  - 'passage: On the basis of the first analysis of transposition measures notified
    by Spain as well as of available information, the Commission considers prima facie
    that Directive 2012/47/EU[1] has been correctly transposed by Spain.

    [1] Directive 2012/47/EU of 14 December 2012 amending Directive 2009/43/EC of
    the European Parliament and of the Council regards the list of defence-related
    products'
  - 'passage: After having carried out a more thorough analysis of the transposition
    measures notified by Spain related to Directive 2012/47/EC, the Commission has
    not found any reason that would justify launching an infringement procedure. In
    line with the provisions of Directive 2009/43/EC on intra-community transfer of
    defence-related products, the list of defence-related products in annex of the
    Directive has to be regularly updated in order to strictly reflect the Common
    Military List of the European Union (CML) annually adopted by the Council. Directive
    2012/47/EC updated annex I of the Directive 2009/43/EC in accordance with the
    CML of 2012. Directive 2012/47/EC has by now been substituted by Directive 2014/108/EU.'
- source_sentence: 'query: EU politicians banned from third countries

    Does the Council maintain a list of which EU politicians are currently banned
    from nations which are not members of the European Union? If so, please provide
    details.'
  sentences:
  - 'passage: Rules and practices on the cooling-off periods applicable to national
    politicians and senior civil servants are of the competence of Member States.
    The European Union''s support competence in the administrative cooperation field
    does not empower it to harmonise those national rules and practices.

    The Commission has its own body of rules (in particular, the Code of Conduct for
    Commissioners and Staff Regulations for EU officials) as regards conflicts of
    interest and post mandate service activities. The Commission''s perception is
    that there is a general understanding amongst administrations that a cooling-off
    period is an efficient tool to cater for the risk of conflicts of interest that
    may stem from post mandate service activities. Such restriction may be imposed
    with due consideration for the principle of proportionality in particular in light
    of the right of every individual to pursue a freely chosen or accepted occupation
    as enshrined in Article 15 of the Charter of Fundamental Rights of the EU.

    However, the Commission does not have any detailed information on the cooling-off
    periods applied in each EU Member States for politicians and senior civil servants.
    Considerable resources would be needed to gather accurate and exhaustive information
    on the rules and practices of every single Member State.'
  - 'passage: For an outermost region such as the Canary Islands the EU Solidarity
    Fund could be activated under two main conditions. Firstly, total direct damage
    caused by the disaster would have to exceed 1% of the regional Gross Domestic
    Product at NUTS2 level. For the Canary Islands this threshold currently amounts
    to EUR 407.2 million. Secondly, the competent national authorities of Spain would
    have to present an application for Solidarity Fund assistance to the Commission
    within 12 weeks of the disaster happening. Until now, the Spanish authorities
    have not submitted an application or indicated that they intend doing so.'
  - 'passage: The information requested by the Honourable Member is not available
    to the Council.'
- source_sentence: 'query: Europe for Citizens - West Midlands

    1. Can the Commission provide details of what funding and initiatives it has provided
    through the Europe for Citizens programme in the West Midlands in the period 2009-2014?

    2. Can the Commission indicate which of these programmes are operating at present?

    3. Can the Commission also provide details of any planned funding to be provided
    in the West Midlands through the Europe for Citizens programme in the next funding
    period?'
  sentences:
  - 'passage: According to FRONTEX, 11 639 migrants illegally entered on the islands
    of Greece via Turkey from 20 March 2016 to 3 July 2016.

    From 4 April 2016 - planned date for the first returns under the statement - to
    6 July 2016, 468 migrants were readmitted to Turkey under the EU-Turkey Statement
    of 18 March 2016.

    More generally, the Commission would like to refer the Honourable Member to its
    reports on the implementation of the EU-Turkey statement.[1]

    [1] In particular the Second Report on the progress made in the implementation
    of the EU-Turkey Statement, COM(2016) 349 final'
  - 'passage: The Commission would like to inform the Honourable Member that the West
    Midlands region (UK) plays an active role in the Europe for Citizens Programme,
    in particular in the area of town-twinning. Between 2009 and 2013, the municipalities
    of Stafford, Leek, Eardisland, Leominster and Worcester situated in the West Midlands
    Region have received grants from the Europe for Citizens Programme for town-twinning
    projects with twin towns, such as Este (Italy), Bad Windsheim (Germany) or Pertuis
    (France).

    The municipality of Leek received three grants for a total amount of EUR 23,964,
    the Stafford and Tarragona Twinning association received a grant of EUR 17,945
    and Eardisland, Leominster and Worcester EUR 5,000 for small-scale town-twinning
    projects. All selected projects were implemented successfully.'
  - 'passage: 1. The funding shown in Annex I was provided through the Europe for
    Citizens programme in the West Midlands in the period 2009-2014

    2. The Europe for Citizens programme 2014-2020 is the successor to the Europe
    for Citizens Programme 2007-2013.

    3. Co-financing under the Europe for Citizens Programme is granted on the basis
    of calls for proposals and the Commission thus cannot know in advance which projects
    will be selected.'
- source_sentence: 'query: Destruction of a Russian plane by the Turkish air force

    The Turkish air force shot down a Russian plane on Tuesday morning. The exact
    circumstances have yet to be established. Nevertheless, Turkey is clearly more
    inclined to take action against Russian or Kurdish forces than against Islamic
    State.

    While a grand coalition is currently being formed to combat Islamic State, Turkey’s
    role in the region is murkier than ever. This confirms that Turkey is not a reliable
    interlocutor, contrary to what Angela Merkel and François Hollande would have
    us believe.

    In the face of this reprehensible act of war, the Council should take up a clear
    position vis-à-vis the Erdoğan government.

    Taking the current situation into account, does the Council intend to reconsider
    the policy on Turkey that it has pursued up to now?'
  sentences:
  - 'passage: Regarding the Council''s policy on Turkey, the Honourable Member is
    invited to refer to the Council''s reply to Written Question [REF].'
  - 'passage: After the downing of a Russian jet plane at the Turkish-Syrian border,
    the EU High Representative Federica Mogherini, in separate phone calls to both
    the Russian and the Turkish Foreign Ministers stressed the need to avoid further
    escalation, preserve the diplomatic talks on Syria and unite forces against Da''esh.

    The EU continues to actively support the UN-led international efforts to find
    a political solution to the Syrian conflict, and is proactively engaging regional
    and international partners. Both Turkey and Russia are part of the International
    Syria Support Group (ISGG) and their role remains critical for the search for
    a solution to the Syria crisis. We expect both Turkey and Russia to actively support
    the ongoing cessation of hostilities and the search for a political solution.'
  - 'passage: Equal access to education irrespective of gender is a priority for the
    EU. This priority is mainstreamed in all EU bilateral cooperation programmes in
    education.

    As regards to the actions taken to improve girls and women access to education,
    the Commission would refer the Honourable Member to its answer to written question

    [REF][1].

    [1] http://www.europarl.europa.eu/plenary/en/parliamentary-questions.html'
pipeline_tag: sentence-similarity
library_name: sentence-transformers
---

# SentenceTransformer based on intfloat/multilingual-e5-base

This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [intfloat/multilingual-e5-base](https://huggingface.co/intfloat/multilingual-e5-base). It maps inputs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, classification, clustering, and more.

## Model Details

### Model Description
- **Model Type:** Sentence Transformer
- **Base model:** [intfloat/multilingual-e5-base](https://huggingface.co/intfloat/multilingual-e5-base) <!-- at revision d128750597153bb5987e10b1c3493a34e5a4502a -->
- **Maximum Sequence Length:** 512 tokens
- **Output Dimensionality:** 768 dimensions
- **Similarity Function:** Cosine Similarity
- **Supported Modality:** Text
<!-- - **Training Dataset:** Unknown -->
<!-- - **Language:** Unknown -->
<!-- - **License:** Unknown -->

### Model Sources

- **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
- **Repository:** [Sentence Transformers on GitHub](https://github.com/huggingface/sentence-transformers)
- **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)

### Full Model Architecture

```
SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'XLMRobertaModel'})
  (1): Pooling({'embedding_dimension': 768, 'pooling_mode': 'mean', 'include_prompt': True})
  (2): Normalize({'module_input_name': 'sentence_embedding', 'module_output_name': 'sentence_embedding'})
)
```

## Usage

### Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

```bash
pip install -U sentence-transformers
```
Then you can load this model and run inference.
```python
from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'query: Destruction of a Russian plane by the Turkish air force\nThe Turkish air force shot down a Russian plane on Tuesday morning. The exact circumstances have yet to be established. Nevertheless, Turkey is clearly more inclined to take action against Russian or Kurdish forces than against Islamic State.\nWhile a grand coalition is currently being formed to combat Islamic State, Turkey’s role in the region is murkier than ever. This confirms that Turkey is not a reliable interlocutor, contrary to what Angela Merkel and François Hollande would have us believe.\nIn the face of this reprehensible act of war, the Council should take up a clear position vis-à-vis the Erdoğan government.\nTaking the current situation into account, does the Council intend to reconsider the policy on Turkey that it has pursued up to now?',
    "passage: Regarding the Council's policy on Turkey, the Honourable Member is invited to refer to the Council's reply to Written Question [REF].",
    "passage: After the downing of a Russian jet plane at the Turkish-Syrian border, the EU High Representative Federica Mogherini, in separate phone calls to both the Russian and the Turkish Foreign Ministers stressed the need to avoid further escalation, preserve the diplomatic talks on Syria and unite forces against Da'esh.\nThe EU continues to actively support the UN-led international efforts to find a political solution to the Syrian conflict, and is proactively engaging regional and international partners. Both Turkey and Russia are part of the International Syria Support Group (ISGG) and their role remains critical for the search for a solution to the Syria crisis. We expect both Turkey and Russia to actively support the ongoing cessation of hostilities and the search for a political solution.",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.8275, 0.4647],
#         [0.8275, 1.0000, 0.2871],
#         [0.4647, 0.2871, 1.0000]])
```
<!--
### Direct Usage (Transformers)

<details><summary>Click to see the direct usage in Transformers</summary>

</details>
-->

<!--
### Downstream Usage (Sentence Transformers)

You can finetune this model on your own dataset.

<details><summary>Click to expand</summary>

</details>
-->

<!--
### Out-of-Scope Use

*List how the model may foreseeably be misused and address what users ought not to do with the model.*
-->

<!--
## Bias, Risks and Limitations

*What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
-->

<!--
### Recommendations

*What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
-->

## Training Details

### Training Dataset

#### Unnamed Dataset

* Size: 59,612 training samples
* Columns: <code>anchor</code>, <code>positive</code>, and <code>negative</code>
* Approximate statistics based on the first 100 samples:
  |          | anchor                                                                               | positive                                                                             | negative                                                                             |
  |:---------|:-------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------|
  | type     | string                                                                               | string                                                                               | string                                                                               |
  | modality | text                                                                                 | text                                                                                 | text                                                                                 |
  | details  | <ul><li>min: 33 tokens</li><li>mean: 238.46 tokens</li><li>max: 512 tokens</li></ul> | <ul><li>min: 23 tokens</li><li>mean: 352.27 tokens</li><li>max: 512 tokens</li></ul> | <ul><li>min: 98 tokens</li><li>mean: 392.77 tokens</li><li>max: 512 tokens</li></ul> |
* Samples:
  | anchor                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       |
  |:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
  | <code>query: The role of merit in unequal societies<br>Inequality is one of the most controversial topics in the European Union, as it has returned to levels seen in the early decades of the last century, albeit with the difference that in today’s Europe the welfare state protects the less privileged.<br>Health care and education are almost free, pensions are often higher than the contributions paid, and unemployment benefit in some countries is higher than in others. Therefore equal opportunities need to be guaranteed for everybody.<br>Equal incentives are needed, rewarding merit and endeavour, especially to protect minorities that find themselves unable to compete despite their endeavour and commitment.<br>The welfare state, which was created for this purpose, must therefore do everything it can to help.<br>Can the Commission say what instruments it considers necessary to eliminate such inequalities and ensure that everybody enjoys equal opportunities?</code>                                                | <code>passage: The European Commission works alongside Member States (MS) to foster equality of opportunity for all and promotes the active inclusion and full participation of vulnerable groups in society.<br>The EU legal framework for equal and anti-discrimination includes the Racial Equality Directive, implementing the principle of equal treatment between persons irrespective of racial or ethnic origin in many areas of social life, and the Employment Equality Directive, which establishes a general framework for equal treatment in employment and occupation.<br>In addition, the Commission:<br> In the context of the European Semester, monitors Member States' policies and provides policy guidance in its annual Country-Specific Recommendations;<br> Supports the implementation of MS' social policies, also through funding[1], e.g. the European Social Fund (ESF), the Youth Employment Initiative, the Fund for European Aid to the Most Deprived (FEAD), the EU Programme for Employment and Social Innovation (EaS...</code> | <code>passage: The Strategy on equality between women and men 2010-2015[1] (Strategy) addresses women's higher poverty risk, particularly within its strand on Equal Economic Independence. One of the key actions is the promotion of gender equality in implementing the Europe 2020 strategy[2]. The Commission’s 2012 White Paper on Pensions provides guidance to Member States on closing the gender pension gap, and the 2013 Social Investment Package urges Member States to improve the social and labour market integration of women.<br>Within the European Semester, the Commission has issued Country Specific Recommendations reflecting these policy priorities. In 2014, 14 Member States received CSRs to: promote female employment through the provision of child-care facilities and the reduction of fiscal disincentives; equalise the pensionable age between women and men; address the poverty risk for older women and ensure equal opportunities for vulnerable individuals. Other actions of the Strategy include th...</code>    |
  | <code>query: EU assistance programme for the Turkish Cypriot community seeking the reunification of Cyprus<br>Since 2006, the European Commission has been approving assistance programmes for the Turkish Cypriot community in Cyprus. The 2021 action programme, totalling EUR 30.7 million, seeks to facilitate the reunification of Cyprus, focusing in particular on the economic integration of the island and closer contacts between the two communities with each other and with the EU. From 2006 to 2021, EUR 623 million was allocated for projects to support confidence-building measures and continue substantial funding for the Committee on Missing Persons and the Technical Committee on Cultural Heritage.<br>However while the aid programme is seeking to facilitate the reunification of the island, the Turkish Cypriot side, with the support and guidance of Turkey, is calling for a 'two-state solution' in Cyprus, contrary to UN Security Council resolutions and EU law. Moreover, a further breach of UN Security C...</code> | <code>passage: The overall objective of the Aid Programme under Council Regulation (EC) No 389/2006[1] is to facilitate the reunification of Cyprus through the implementation of projects under the six objectives outlined in the Regulation.<br>An external evaluation, which will be published shortly, has examined the Aid Programme in the 2013-2018 period against six evaluation criteria: relevance, effectiveness, efficiency, coherence, EU added value and sustainability. The evaluation assessed a sample of 22 interventions that covered all six objectives of the Aid Programme, corresponding to around 60% of the total operational budget of EUR 199 million. The evaluation concluded that the Aid Programme remains relevant due to the persisting gap in socio-economic development and the absence of a comprehensive settlement. It also found that the programme was generally able to achieve the intended results despite, in some cases, delays and sustainability issues due to capacity problems on the beneficia...</code>          | <code>passage: The best way forward to resolve the Cyprus problem is through a resumption of formal negotiations of a comprehensive settlement. The EU remains fully committed to the comprehensive settlement based on a bicommunal, bizonal federation with political equality, in accordance with the relevant United Nations (UN) Security Council Resolutions and the principles on which the EU is founded[1].<br>The EU will continue to play an active role in supporting the process. The Commission implements an aid programme[2] to facilitate the reunification of Cyprus by encouraging the economic development of the Turkish Cypriot community with particular emphasis on the economic integration of the island, on improving contacts between the two communities and with the EU, and on preparation for the EU acquis. Between 2006 and now, about EUR 600 million have been allocated to this programme.<br>The commitment of all parties involved to creating the conditions for the comprehensive settlement remains crucia...</code> |
  | <code>query: VP/HR - Iran compulsory dress code<br>In December 2017, a recent wave of protests were sparked when a woman standing on Teheran’s Revolution Street posted a picture on social media of herself without a headscarf. Her peaceful protest generated many others. Including the original protester, numerous other women were arrested and charged with things like ‘openly committing a sinful act’ and ‘encouraging immorality or prostitution’, which can earn them up to ten years in prison if they are convicted. Iran’s strict dress code violates women’s right to freedom of expression and religion.<br>What can be done to encourage Iran to be more lenient with the dress code?</code>                                                                                                                                                                                                                                                                                                                                                | <code>passage: The wave of protests in December 2017 which were driven mostly by economic and social grievances, have sparked a debate in the country on several issues including on the compulsory Islamic dress code for women. President Rouhani appealed for greater personal freedoms at the annual celebration of the Islamic Revolution on 11 February 2018 and spoke in favor of organising referendums on certain topics of debate in Iran. Also the Vice-President for Women and Family Affairs, Masoumeh Ebtekar, has emphasized the need for a dialogue with the youth on the mandatory hijab and to use methods that are lawful instead of using police force.<br>The EU attaches high importance to human rights and fundamental freedoms in Iran. During the latest round of the High Level Dialogue with Iran, which took place in November 2017 in Tehran, the rights of women were one of the key topics discussed. In this context, the EU has also met with the Vice-President for Women and Family Affairs, Masoumeh Ebtekar...</code>          | <code>passage: The waves of protests in Iran which have taken place in Iran since the beginning of 2018 have sparked a debate in the country on several issues including on the compulsory Islamic dress code for women. President Rouhani appealed for greater personal freedoms at the annual celebration of the Islamic Revolution on 11 February 2018 and spoke in favor of organising referendums on certain topics of debate in Iran. Also the Vice-President for Women and Family Affairs, Masoumeh Ebtekar, has emphasized the need for a dialogue with the youth on the mandatory hijab and to use methods that are lawful instead of using police force.<br>The EU attaches high importance to human rights and fundamental freedoms in Iran. During the latest round of the High Level Dialogue with Iran, which took place in November 2017 in Tehran, the rights of women were one of the key topics discussed. In this context, the EU has also met with the Vice-President for Women and Family Affairs, Masoumeh Ebtekar and agre...</code>    |
* Loss: [<code>CachedMultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cachedmultiplenegativesrankingloss) with these parameters:
  ```json
  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "mini_batch_size": 8,
      "mini_batch_num_tokens": null,
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  ```

### Training Hyperparameters
#### Non-Default Hyperparameters

- `per_device_train_batch_size`: 32
- `num_train_epochs`: 1
- `learning_rate`: 2e-05
- `warmup_steps`: 0.1
- `fp16`: True
- `seed`: 14
- `dataloader_num_workers`: 2
- `batch_sampler`: no_duplicates

#### All Hyperparameters
<details><summary>Click to expand</summary>

- `per_device_train_batch_size`: 32
- `num_train_epochs`: 1
- `max_steps`: -1
- `learning_rate`: 2e-05
- `lr_scheduler_type`: linear
- `lr_scheduler_kwargs`: None
- `warmup_steps`: 0.1
- `optim`: adamw_torch_fused
- `optim_args`: None
- `weight_decay`: 0.0
- `adam_beta1`: 0.9
- `adam_beta2`: 0.999
- `adam_epsilon`: 1e-08
- `optim_target_modules`: None
- `gradient_accumulation_steps`: 1
- `average_tokens_across_devices`: True
- `max_grad_norm`: 1.0
- `label_smoothing_factor`: 0.0
- `bf16`: False
- `fp16`: True
- `bf16_full_eval`: False
- `fp16_full_eval`: False
- `tf32`: None
- `gradient_checkpointing`: False
- `gradient_checkpointing_kwargs`: None
- `torch_compile`: False
- `torch_compile_backend`: None
- `torch_compile_mode`: None
- `use_liger_kernel`: False
- `liger_kernel_config`: None
- `use_cache`: False
- `neftune_noise_alpha`: None
- `torch_empty_cache_steps`: None
- `auto_find_batch_size`: False
- `log_on_each_node`: True
- `logging_nan_inf_filter`: True
- `include_num_input_tokens_seen`: no
- `log_level`: passive
- `log_level_replica`: warning
- `disable_tqdm`: False
- `project`: huggingface
- `trackio_space_id`: None
- `trackio_bucket_id`: None
- `trackio_static_space_id`: None
- `per_device_eval_batch_size`: 8
- `prediction_loss_only`: True
- `eval_on_start`: False
- `eval_do_concat_batches`: True
- `eval_use_gather_object`: False
- `eval_accumulation_steps`: None
- `include_for_metrics`: []
- `batch_eval_metrics`: False
- `save_only_model`: False
- `save_on_each_node`: False
- `enable_jit_checkpoint`: False
- `push_to_hub`: False
- `hub_private_repo`: None
- `hub_model_id`: None
- `hub_strategy`: every_save
- `hub_always_push`: False
- `hub_revision`: None
- `load_best_model_at_end`: False
- `ignore_data_skip`: False
- `restore_callback_states_from_checkpoint`: False
- `full_determinism`: False
- `seed`: 14
- `data_seed`: None
- `use_cpu`: False
- `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
- `parallelism_config`: None
- `dataloader_drop_last`: False
- `dataloader_num_workers`: 2
- `dataloader_pin_memory`: True
- `dataloader_persistent_workers`: False
- `dataloader_prefetch_factor`: None
- `dataloader_multiprocessing_context`: None
- `dataloader_in_order`: True
- `remove_unused_columns`: True
- `label_names`: None
- `train_sampling_strategy`: random
- `length_column_name`: length
- `ddp_find_unused_parameters`: None
- `ddp_bucket_cap_mb`: None
- `ddp_broadcast_buffers`: False
- `ddp_static_graph`: None
- `ddp_backend`: None
- `ddp_timeout`: 1800
- `fsdp`: None
- `fsdp_config`: None
- `deepspeed`: None
- `debug`: []
- `skip_memory_metrics`: True
- `do_predict`: False
- `resume_from_checkpoint`: None
- `local_rank`: -1
- `prompts`: None
- `batch_sampler`: no_duplicates
- `multi_dataset_batch_sampler`: proportional
- `router_mapping`: {}
- `learning_rate_mapping`: {}
- `warmup_ratio`: None

</details>

### Training Logs
| Epoch  | Step | Training Loss |
|:------:|:----:|:-------------:|
| 0.0268 | 50   | 2.2481        |
| 0.0537 | 100  | 0.6475        |
| 0.0805 | 150  | 0.4511        |
| 0.1074 | 200  | 0.4033        |
| 0.1342 | 250  | 0.3842        |
| 0.1610 | 300  | 0.3371        |
| 0.1879 | 350  | 0.3596        |
| 0.2147 | 400  | 0.3460        |
| 0.2415 | 450  | 0.3164        |
| 0.2684 | 500  | 0.3371        |
| 0.2952 | 550  | 0.2960        |
| 0.3221 | 600  | 0.3085        |
| 0.3489 | 650  | 0.3128        |
| 0.3757 | 700  | 0.2875        |
| 0.4026 | 750  | 0.2928        |
| 0.4294 | 800  | 0.2769        |
| 0.4563 | 850  | 0.2924        |
| 0.4831 | 900  | 0.2833        |
| 0.5099 | 950  | 0.2829        |
| 0.5368 | 1000 | 0.3053        |
| 0.5636 | 1050 | 0.2766        |
| 0.5904 | 1100 | 0.2984        |
| 0.6173 | 1150 | 0.2698        |
| 0.6441 | 1200 | 0.2525        |
| 0.6710 | 1250 | 0.2651        |
| 0.6978 | 1300 | 0.2811        |
| 0.7246 | 1350 | 0.2733        |
| 0.7515 | 1400 | 0.2428        |
| 0.7783 | 1450 | 0.2828        |
| 0.8052 | 1500 | 0.2765        |
| 0.8320 | 1550 | 0.2523        |
| 0.8588 | 1600 | 0.2672        |
| 0.8857 | 1650 | 0.2838        |
| 0.9125 | 1700 | 0.2467        |
| 0.9393 | 1750 | 0.2746        |
| 0.9662 | 1800 | 0.2511        |
| 0.9930 | 1850 | 0.2642        |


### Training Time
- **Training**: 16.0 minutes

### Framework Versions
- Python: 3.14.4
- Sentence Transformers: 6.1.0
- Transformers: 5.20.2
- PyTorch: 2.14.1+cu130
- Accelerate: 1.15.0
- Datasets: 5.1.0
- Tokenizers: 0.23.3

## Additional Resources

- [Training and Finetuning Embedding Models with Sentence Transformers](https://huggingface.co/blog/train-sentence-transformers): the end-to-end guide for training or finetuning Sentence Transformer models.
- [Introduction to Matryoshka Embedding Models](https://huggingface.co/blog/matryoshka): variable-size embeddings that can be truncated with minimal quality loss.
- [Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval](https://huggingface.co/blog/embedding-quantization): post-training compression of embedding vectors.
- [Multimodal Embedding & Reranker Models with Sentence Transformers](https://huggingface.co/blog/multimodal-sentence-transformers): use text, image, audio, and video models through the same API.
- [Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers](https://huggingface.co/blog/train-multimodal-sentence-transformers): train multimodal embedding models, with a Visual Document Retrieval walkthrough.

## Citation

### BibTeX

#### Sentence Transformers
```bibtex
@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}
```

#### CachedMultipleNegativesRankingLoss
```bibtex
@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}
```

#### MultipleNegativesRankingLoss
```bibtex
@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}
```

<!--
## Glossary

*Clearly define terms in order to be accessible across audiences.*
-->

<!--
## Model Card Authors

*Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
-->

<!--
## Model Card Contact

*Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
-->