Add files using upload-large-folder tool

v1.0/experiments/models/me5-base-ft/1_Pooling/config.json

@@ -0,0 +1,5 @@
+{
+    "embedding_dimension": 768,
+    "pooling_mode": "mean",
+    "include_prompt": true
+}

v1.0/experiments/models/me5-base-ft/2_Normalize/config.json

@@ -0,0 +1,4 @@
+{
+    "module_input_name": "sentence_embedding",
+    "module_output_name": "sentence_embedding"
+}

v1.0/experiments/models/me5-base-ft/README.md

@@ -0,0 +1,550 @@
+---
+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: Spanish failure to submit monthly reports regarding Gibraltar
+
+    What action does the Commission propose to take against Spain following its failure
+    to submit monthly reports describing the risks, actions and results of the situation
+    at its land frontier with Gibraltar, as requested by the Commission in its letter
+    of 30 July 2014?'
+  sentences:
+  - 'passage: Upon request from the Commission, the Spanish authorities have submitted
+    on 20 February the reports on the checks carried out as well as the number of
+    seizure, administrative offences or crimes related to tobacco smuggling. These
+    reports cover the period May - December 2014.
+
+    The Commission is continuing the dialogue with Spanish authorities in order to
+    ensure that the relevant data and reports are being timely submitted.'
+  - 'passage: The Spanish authorities have submitted the Action protocol establishing
+    guidelines on risk analysis at its land border with Gibraltar on 2 February 2015.
+    The monthly reports to which the Honourable Member is referring have not been
+    provided by Spain so far.'
+  - 'passage: The proposal referred to by the Honourable Member is being discussed
+    within the Council. The Council is not in a position to foresee the outcome or
+    the duration of the discussions.
+
+    The latest progress report is set out in 13877/1/15 REV 1.'
+- source_sentence: 'query: UK requests for assistance from the EU Food Fraud Network
+
+    On how many occasions in the last year have the UK Food Standards Agency and Food
+    Standards Scotland requested assistance through the EU Food Fraud Network, and
+    in relation to what specific matters?'
+  sentences:
+  - 'passage: In 2017, the UK Food Standards Agency and Food Standards Scotland requested
+    assistance through the EU Food Fraud Network (FFN) on 10 occasions. The subjects
+    of these requests were mislabelling, online sale of substances unfit for human
+    consumption, absent/falsified documentation, unapproved treatment with pesticides,
+    organic certification fraud, product/ingredient substitution and suspected illegal
+    harvesting.'
+  - 'passage: Pursuant to Article 153(5) TFEU, the EU has no competence in matters
+    of pay. Member States'' national law governs situations such as the one described
+    by the Honourable Member.'
+  - 'passage: In 2018, the United Kingdom Food Standards Agency and Food Standards
+    Scotland requested assistance through the EU Food Fraud Network on seven occasions.
+    These requests related to suspicions of mislabelling, addition of undeclared/unauthorized
+    substances, nutrition/health claims, counterfeit goods (trademarks), product replacement/substitution
+    and online sale of substances unfit for human consumption.'
+- source_sentence: 'query: Coding for the sharing economy
+
+    In its answer to Written Question [REF], the Commission was unable to answer the
+    question regarding the amount of money spent in the sharing economy by officials
+    on mission because sufficient coding was not available. Beyond the ability to
+    answer this question, does the Commission believe that it may be useful to code
+    for such expenses?'
+  sentences:
+  - 'passage: The Commission and its staff are bound to comply with the rules laid
+    down by the Financial Regulation, the Staff Regulations and its implementing rules,
+    such as the Guide to missions. Therefore, the Commission''s staff arrange their
+    trips according to the rule of cost-effectiveness. In this context, whenever the
+    requirements of the Guide to missions are met, the use of sharing economy services
+    is permitted.
+
+    The possibility to enable coding for such expenses in the IT tool could be envisaged
+    in the context of broader legal and budgetary feasibility analysis.'
+  - 'passage: The Commission regrets to inform the Honourable Member that it does
+    not have sufficient information to be able to answer the question of the amount
+    of money spent in the sharing economy by officials on mission. The IT systems
+    used to manage and reimburse expenses incurred whilst on official business focuses
+    on amount, type and proof of the expenses and does not provide for encoding of
+    precise references to the companies concerned. However, Commission Staff are allowed
+    to use sharing economy services, if they can provide an invoice, which is the
+    case for an increasing number of those companies.'
+  - 'passage: The Commission refers the Honourable Member to its reply to question
+    [REF].'
+- source_sentence: 'query: Compensation for costs
+
+    In its answer to question [REF], the Commission said that, in accordance with
+    the EU’s ETS Directive and the relevant state-aid guidelines, as from 1 January
+    2013 industries vulnerable to carbon leakage may be granted aid compensating them
+    for the knock-on effects on electricity prices of ETS-related CO2 emissions costs.
+
+    Over the years, various Member States have adopted different aid schemes. Could
+    the Commission give details of the countries that have approved compensation schemes
+    since 2013 and the maximum amounts authorised per year (and if possible the amounts
+    actually granted) since 2013.'
+  sentences:
+  - 'passage: The Commission will kindly refer the Honourable Member to written answer
+    [REF][1].
+
+    [1] http://www.europarl.europa.eu/plenary/en/parliamentary-questions.html'
+  - 'passage: The Table in annex presents the list of Member States (and Norway) having
+    introduced an EU Emissions Trading System compensation scheme during the Phase
+    III of the EU ETS(2013/2020).'
+  - 'passage: The Commission has approved six State aid measures compensating energy-intensive
+    industries at risk of carbon leakage for indirect CO2 emission costs due to the
+    EU Emissions Trading System (ETS), passed on in electricity prices. Under the
+    EU ETS Directive and relevant State aid guidelines, aid may be granted for costs
+    incurred on or after 1 January 2013. The table in the Annex summarises the available
+    data on aid amounts for each of these measures.
+
+    The Commission''s Guidelines on State Aid for Environmental Protection and Energy
+    2014-2020 allow Member States to relieve some EU energy-intensive sectors particularly
+    exposed to international competition from some of the cost of financing support
+    for renewable energy. These Guidelines were adopted on 9 April 2014. Prior to
+    their adoption, no such aid was approved by the Commission. Annual reports are
+    therefore not yet available on such measures.
+
+    The Commission will no later than 16 March 2014 inform Parliament of its reply
+    to Resolution 2014/2976(RSP) of 17 December 2014 in accordance with usual procedures
+    regarding non-legislative resolutions.'
+- source_sentence: 'query: Restructuring of HSH Nordbank
+
+    According to media reports, the Commission has set up a ''case team'' headed by
+    Commissioner Vestager to examine the situation regarding HSH Nordbank[1]. In addition,
+    Bain & Company, a global management consulting firm, has also produced a study
+    on ways of rationalizing HSH Nordbank, for example the sale of sale of non-performing
+    loans of between EUR 3 billion and EUR 14 billion to the states of Hamburg and
+    Schleswig-Holstein.
+
+    1. What specific rationalisation measures have been recommended by the states
+    of Hamburg and Schleswig-Holstein and by HSH Nordbank itself?
+
+    2. What precise conditions must be met by HSH to receive a clean bill of health
+    from the ECB Single Supervisory Mechanism and at what point will it be faced with
+    the threat of liquidation?
+
+    3. What conditions are being stipulated by the Commission for approval of aid
+    (further guarantee increase from EUR 7 billion to EUR 10 billion) and when are
+    the state aid proceedings likely to be over?
+
+    [1] https://www.ndr.de/nachrichten/schleswig-holstein/HSH-Nordbank-hofft-auf-Milliarden-Finanzspritze,hshnordbank736.html'
+  sentences:
+  - 'passage: The Treaties are the basis for a united, open and stronger Europe, the
+    realisation of which is the guiding principle of the Commission''s day-to-day
+    work.
+
+    The Commission has set out its plans for achieving a deep and genuine EMU in its
+    Blueprint launched in November 2012. The Blueprint provides a comprehensive vision
+    for a deep and genuine EMU conducive to a strong and stable architecture in the
+    financial, fiscal, economic and political domains to underpin stability and prosperity
+    in the future..
+
+    Moreover, President Barroso has offered some further contributions to the ongoing
+    debate on the future of Europe and Political Union in his speech of 8 May 2014
+    at Berlin''s Humboldt University.[1]
+
+    [1] http://europa.eu/rapid/press-release_SPEECH-14-355_en.htm'
+  - 'passage: According to the October 2015 agreement in principle between the Commission
+    and the Ministerpräsidenten of Hamburg and Schleswig-Holstein, HSH Nordbank can
+    sell up to EUR 6.2 billion of assets (Exposure at Default) to the two Länder.
+    In order to ensure that the transfer does not entail aid to the bank, the sale
+    of assets must take place at market values.
+
+    The Commission''s independent valuation experts have undertaken the determination
+    of the market value for the relevant assets for the final decision on HSH Nordbank
+    adopted on 2 May 2016. Before the publication of the non-confidential version
+    of the decision, the Commission is not in a position to provide those values.'
+  - 'passage: The State aid investigation in the case of HSH Nordbank is currently
+    ongoing. The Commission is not in a position to comment on ongoing cases and is
+    bound in this respect by the applicable rules and regulations.'
+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: Restructuring of HSH Nordbank\nAccording to media reports, the Commission has set up a 'case team' headed by Commissioner Vestager to examine the situation regarding HSH Nordbank[1]. In addition, Bain & Company, a global management consulting firm, has also produced a study on ways of rationalizing HSH Nordbank, for example the sale of sale of non-performing loans of between EUR 3 billion and EUR 14 billion to the states of Hamburg and Schleswig-Holstein.\n1. What specific rationalisation measures have been recommended by the states of Hamburg and Schleswig-Holstein and by HSH Nordbank itself?\n2. What precise conditions must be met by HSH to receive a clean bill of health from the ECB Single Supervisory Mechanism and at what point will it be faced with the threat of liquidation?\n3. What conditions are being stipulated by the Commission for approval of aid (further guarantee increase from EUR 7 billion to EUR 10 billion) and when are the state aid proceedings likely to be over?\n[1] https://www.ndr.de/nachrichten/schleswig-holstein/HSH-Nordbank-hofft-auf-Milliarden-Finanzspritze,hshnordbank736.html",
+    'passage: The State aid investigation in the case of HSH Nordbank is currently ongoing. The Commission is not in a position to comment on ongoing cases and is bound in this respect by the applicable rules and regulations.',
+    "passage: According to the October 2015 agreement in principle between the Commission and the Ministerpräsidenten of Hamburg and Schleswig-Holstein, HSH Nordbank can sell up to EUR 6.2 billion of assets (Exposure at Default) to the two Länder. In order to ensure that the transfer does not entail aid to the bank, the sale of assets must take place at market values.\nThe Commission's independent valuation experts have undertaken the determination of the market value for the relevant assets for the final decision on HSH Nordbank adopted on 2 May 2016. Before the publication of the non-confidential version of the decision, the Commission is not in a position to provide those values.",
+]
+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, 0.4984, 0.3288],
+#         [0.4984, 1.0000, 0.4098],
+#         [0.3288, 0.4098, 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: 47 tokens</li><li>mean: 225.93 tokens</li><li>max: 403 tokens</li></ul> | <ul><li>min: 53 tokens</li><li>mean: 320.8 tokens</li><li>max: 512 tokens</li></ul> | <ul><li>min: 147 tokens</li><li>mean: 401.79 tokens</li><li>max: 512 tokens</li></ul> |
+* Samples:
+  | anchor                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                          |
+  |:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+  | <code>query: Co-financing of projects<br>The European rules of governance and the pressure of debt markets in which the ECB has refused to intervene have caused many Member States to make drastic adjustments to their budgets. As a result of the rules on co-financing, the repercussions of these cuts snowball when they prevent access to EU funds in projects of mutual interest.<br>Given the above, will the Commission reduce the co-financing rates for projects by Member States' public administrations, particularly for social cohesion and convergence projects, such as those financed under the European Regional Development Fund (ERDF) or the European Social Fund (ESF)?<br>Will the Commission extend deadlines so that Member States can pay their contributions, thus postponing the payments of Member States with serious budget difficulties?<br>Finally, given that the Commission is creating new initiatives - such as the European Fund for Strategic Investments (EFSI) - which are aimed exclusively at the mobilisatio...</code> | <code>passage: 1. Regulation (EU) N° 1297/2013 of the European Parliament and the Council of 11 December 2013[1] allows payments under the 2007-2013 programmes to be increased by an amount corresponding to 10 % above the co-financing rate applicable to each priority, but not exceeding 100%, where Member States benefit from one of the EU financial assistance mechanisms referred to in the regulation. This increase applies to the amount of eligible expenditure newly declared in each certified statement of expenditure submitted until the end of the programming period.<br>2. The final date of eligibility of expenditure is 31 December 2015 under the 2007-2013 period and this deadline will not be extended. Eligible expenditure means expenditure incurred and paid by the beneficiary. Council Regulation 1083/2006 of 11 July 2006[2] does not specify any final date for the reimbursement to the beneficiaries of the public contribution but states that it should be paid "as quickly as possible".<br>3. The Commis...</code> | <code>passage: Retrospective projects are eligible for European Regional Development Fund (ERDF) funding provided that they comply with the rules set in the EU legal framework. In the case of Italy, there are specific national provisions on retrospective projects established in the national strategic framework under the 2007-2013 period. At present, there are no specific data on the share of funding used for retrospective projects since there is no obligation to report it in the final implementation report 2007-2013. However, the closure procedure is ongoing. The Commission formulated a series of questions, including a specific one on the amount allocated to retrospective projects. The Commission is waiting for a reply.<br>At closure, the ERDF managing authority reported 9 phased projects to be carried over in the 2014-2020 period. The total expenditure allocated to these projects under the 2007-2013 programme totals EUR 6.7 million.<br>At closure, there is no national co-financing transferred fro...</code>    |
+  | <code>query: EU-Ukraine Association Agreement<br>Does the Commission accept that, given the current situation in Ukraine, the signing of the EU-Ukraine Association Agreement, which aims to ‘deepen political association’ (Article 4(2)(a)), might actually further exacerbate political tensions within that country?</code>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      | <code>passage: The EU has deeply engaged in seeking a peaceful settlement of the crisis in Ukraine. We have to find ways to explain to Russia that respect can be regained if we work jointly on a vision of cooperation on the European continent without dividing lines. The Association Agreement is not meant to replace - or intrude in - Ukraine's relations with any neighbour, but we have taken note of the public support for closer association and integration with Europe. Throughout the year, the European Union and Ukraine have engaged with Russia in order to discuss the economic effects on the Russian economy of the implementation of the AA/DCFTA with a view to clarify concerns and - where appropriate - identify solutions.<br>A concrete result of these consultations is the agreement reached at the trilateral ministerial meeting on 12 September between the EU, Ukraine and Russia. Consequently, the Commission has proposed to the Council to delay until 31 December 2015 the provisional application of t...</code>    | <code>passage: The EU notes evidence of continued support given to the separatists by Russia, which underlines Russia's responsibility in destabilizing the situation in eastern Ukraine.<br>Ukraine is currently engaged in a broad and complex reform process aiming to respond to its citizens's aspirations for a democratic, secure and prosperous country.<br>The European Union has put in place an unprecedented package of measures in support of Ukraine's reform steps, including political support, financial, technical and economic assistance, and strategic advice. The EU maintains close coordination with the Ukrainian authorities to ensure the commitments undertaken within the Association Agreement framework are fully implemented. The EU-Ukraine Association Agenda is in this context the main political tool for the implementation and monitoring of the Association Agreement.<br>The EU has also joined the international efforts and stepped-up its humanitarian assistance to the people affected by the conflict in...</code> |
+  | <code>query: CBRN preparedness and protection measures<br>On 17 January 2018, the World Economic Forum published on its latest annual Global Risks Report. The likelihood that weapons of mass destruction might be used has again increased compared with last year’s assessment. International reports provide more and more reasons to believe that the threat of the deliberate use of CBRN weapons is becoming more plausible. In October 2017, the Commission adopted an Action Plan to enhance preparedness against chemical, biological, radiological and nuclear security risks, containing several proposals aimed at supporting Member States in better defending EU citizens against terrorist threats. However, the level of risk awareness and, therefore, preparedness in the event of a CBRN incident differs widely between all 28 Member States.<br>1. Knowing that only a few Member States would be sufficiently protected in the event of a CBRN attack, and knowing that such attacks typically have a cross-border impact, wh...</code>       | <code>passage: Protection of citizens and infrastructures remains primarily responsibility of Member States of the European Union. Nevertheless in the Action Plan to enhance preparedness against chemical, biological, radiological and nuclear security risks[1] adopted on 18 October 2017, the Commission proposed several initiatives focusing on sharing of expertise and resources among Member States aiming at better preparedness of the EU as a whole for the chemical, biological, radiological and nuclear (CBRN) terrorist threats. Among others the Commission proposed development of a common EU CBRN training curriculum, boosting of cross-border trainings and exercises and encouraging Member States to commit more capacities to the European Emergency Response Capacity (EERC) of the Union's Civil Protection Mechanism. These - and other initiatives announced in the Action Plan - should over time result in an enhanced preparedness of the EU for the CBRN terrorist threats.<br>[1] COM(2017) 610 final</code>               | <code>passage: Internal security, including the protection of citizens, residents and infrastructures is the primary responsibility of the Member States of the European Union.<br>Nevertheless, in the Action Plan to enhance preparedness against chemical, biological, radiological and nuclear (CBRN) security risks[1] adopted on 18 October 2017, the Commission proposed several initiatives focusing on sharing of expertise and resources among Member States aiming at better preparedness of the EU as a whole against threats from CBRN agents. Among others, the Commission proposed the development of a common EU CBRN training curriculum or boosting of cross-border trainings. The initiatives announced in the Action Plan will over time result in an enhanced preparedness of the EU against CBRN security risks as well as a higher security level of EU citizens and residents.<br>Moreover, following the Salisbury attack and at the invitation of the European Council[2], the Commission and the High Representative prop...</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`: 13
+- `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`: 13
+- `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   | 2.2747        |
+| 0.0537 | 100  | 0.6591        |
+| 0.0805 | 150  | 0.4337        |
+| 0.1074 | 200  | 0.4104        |
+| 0.1342 | 250  | 0.3719        |
+| 0.1610 | 300  | 0.3429        |
+| 0.1879 | 350  | 0.3468        |
+| 0.2147 | 400  | 0.3335        |
+| 0.2415 | 450  | 0.3253        |
+| 0.2684 | 500  | 0.3086        |
+| 0.2952 | 550  | 0.3005        |
+| 0.3221 | 600  | 0.2995        |
+| 0.3489 | 650  | 0.3106        |
+| 0.3757 | 700  | 0.2956        |
+| 0.4026 | 750  | 0.2878        |
+| 0.4294 | 800  | 0.3074        |
+| 0.4563 | 850  | 0.2837        |
+| 0.4831 | 900  | 0.2683        |
+| 0.5099 | 950  | 0.2908        |
+| 0.5368 | 1000 | 0.2658        |
+| 0.5636 | 1050 | 0.2707        |
+| 0.5904 | 1100 | 0.2456        |
+| 0.6173 | 1150 | 0.2702        |
+| 0.6441 | 1200 | 0.2478        |
+| 0.6710 | 1250 | 0.2650        |
+| 0.6978 | 1300 | 0.2605        |
+| 0.7246 | 1350 | 0.2655        |
+| 0.7515 | 1400 | 0.2494        |
+| 0.7783 | 1450 | 0.2435        |
+| 0.8052 | 1500 | 0.2388        |
+| 0.8320 | 1550 | 0.2166        |
+| 0.8588 | 1600 | 0.2327        |
+| 0.8857 | 1650 | 0.2652        |
+| 0.9125 | 1700 | 0.2320        |
+| 0.9393 | 1750 | 0.2273        |
+| 0.9662 | 1800 | 0.2311        |
+| 0.9930 | 1850 | 0.2642        |
+
+
+### Training Time
+- **Training**: 16.1 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.*
+-->

v1.0/experiments/models/me5-base-ft/config.json

@@ -0,0 +1,30 @@
+{
+  "add_cross_attention": false,
+  "architectures": [
+    "XLMRobertaModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "is_decoder": false,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "xlm-roberta",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "output_past": true,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "tie_word_embeddings": true,
+  "transformers_version": "5.20.2",
+  "type_vocab_size": 1,
+  "use_cache": false,
+  "vocab_size": 250002
+}

v1.0/experiments/models/me5-base-ft/config_sentence_transformers.json

@@ -0,0 +1,14 @@
+{
+  "__version__": {
+    "pytorch": "2.14.1+cu130",
+    "sentence_transformers": "6.1.0",
+    "transformers": "5.20.2"
+  },
+  "default_prompt_name": null,
+  "model_type": "SentenceTransformer",
+  "prompts": {
+    "document": "",
+    "query": ""
+  },
+  "similarity_fn_name": "cosine"
+}

v1.0/experiments/models/me5-base-ft/modules.json

@@ -0,0 +1,20 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.base.modules.transformer.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.sentence_transformer.modules.pooling.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.base.modules.normalize.Normalize"
+  }
+]

v1.0/experiments/models/me5-base-ft/parlir_train_meta.json

@@ -0,0 +1,12 @@
+{
+ "base": "intfloat/multilingual-e5-base",
+ "n_pairs": 59612,
+ "epochs": 1,
+ "batch_size": 32,
+ "lr": 2e-05,
+ "max_len": 512,
+ "seed": 13,
+ "hard_negatives": true,
+ "min_year": 0,
+ "build": "data/build/v1.0"
+}

v1.0/experiments/models/me5-base-ft/sentence_bert_config.json

@@ -0,0 +1,10 @@
+{
+    "transformer_task": "feature-extraction",
+    "modality_config": {
+        "text": {
+            "method": "forward",
+            "method_output_name": "last_hidden_state"
+        }
+    },
+    "module_output_name": "token_embeddings"
+}

v1.0/experiments/models/me5-base-ft/tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "add_prefix_space": true,
+  "backend": "tokenizers",
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 512,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "tokenizer_class": "XLMRobertaTokenizer",
+  "unk_token": "<unk>"
+}

v1.0/experiments/models/me5-base-ft2019/1_Pooling/config.json

@@ -0,0 +1,5 @@
+{
+    "embedding_dimension": 768,
+    "pooling_mode": "mean",
+    "include_prompt": true
+}

v1.0/experiments/models/me5-base-ft2019/2_Normalize/config.json

@@ -0,0 +1,4 @@
+{
+    "module_input_name": "sentence_embedding",
+    "module_output_name": "sentence_embedding"
+}

v1.0/experiments/models/me5-base-ft2019/README.md

@@ -0,0 +1,590 @@
+---
+tags:
+- sentence-transformers
+- sentence-similarity
+- feature-extraction
+- dense
+- generated_from_trainer
+- dataset_size:16873
+- loss:CachedMultipleNegativesRankingLoss
+- loss:MultipleNegativesRankingLoss
+base_model: intfloat/multilingual-e5-base
+widget:
+- source_sentence: 'query: Council Note 7497/20 on demographic challenges in the EU
+
+    Can the Council indicate which delegations have approved Council Note 7497/20
+    on demographic challenges in the EU?'
+  sentences:
+  - 'passage: The interinstitutional negotiations on the proposals for the next multiannual
+    financial framework and the sectoral regulations presented by the European Commission
+    in May and June 2018 are ongoing. The European Council in June 2019 reconfirmed
+    its aim for an agreement before the end of the year 2019. The European Commission
+    remains fully committed to this aim.
+
+    However, in the event that no Council regulation determining a new financial framework
+    would be adopted by the end of the current financial framework, Article 312 (4)
+    of the Treaty on the Functioning of the European Union would apply. This Article
+    provides that the ceilings and other provisions corresponding to the last year
+    of the previous framework shall be extended until such time as that act is adopted.'
+  - 'passage: The note referred to by the Honourable Member was an intermediary step
+    towards the approval of the Council Conclusions on Demographic Challenges - the
+    Way Ahead. The final wording of these Council Conclusions[1] was approved by consensus.
+
+    ______________________
+
+    [1] ST 8668/20'
+  - 'passage: It is not for the Council to explain terms commonly used in demographic
+    studies or policy documents[1]. As for the second and third questions, the Honourable
+    Member is reminded that it is not Council policy to answer hypothetical questions.
+
+    The Honourable Member is also reminded that document 7497/20 of 23 April 2020
+    was an intermediary step towards the approval of the Council conclusions on ‘Demographic
+
+    Challenges - The Way Ahead’[2] on 8 June 2020.
+
+    [1] See for instance the Commission communication of 12 October 2006, COM(2006)
+    0571 final: ‘The demographic future of Europe - from challenge to opportunity’.
+
+    [2] ST 8668/20.'
+- source_sentence: 'query: Consultancy contracts
+
+    Have consultancy contracts been concluded with external consultants under the
+    new Commission?
+
+    If so, with whom and for what amount?'
+  sentences:
+  - 'passage: At this stage it is not possible to reply in detail to the question
+    whether the Commission concluded any consultancy contracts with external consultants
+    between 1 May 2020 and 31 August 2020. The contract category information (e.g.
+    consultancy) becomes available through contract registration in the central accounting
+    IT-system. Contract registration is mandatory when proceeding to the first payment,
+    which can occur several months after the contract has been concluded.
+
+    To ensure the correctness and completeness of contracts, the Commission proceeds
+    to data quality verifications before publishing beneficiaries and related contracts
+    on the Financial Transparency System website (FTS database)[1].
+
+    The publication of 2020 data on the FTS website is foreseen by June 2021 and contracts
+    related to consultancy (name of the contractor and amount) will then become publicly
+    available.
+
+    Information on consultancy contracts for the financial years 2018 and 2019 was
+    provided in replies to the written questions [REF] and [REF].
+
+    [1] https://ec.europa.eu/budget/fts/index_en.htm'
+  - 'passage: In annex 1, the Honourable Member will find all consultancy contracts,
+    together with the contractor(s) name (in column A) and the contracted amount (in
+    column B), that the new European Commission signed from 1 December 2019 to 30
+    April 2020.'
+  - 'passage: The Commission is currently engaged in constructive technical discussions
+    with the Italian authorities in order to analyse, from a State aid and public
+    procurement perspective, the measures for the completion of the works on the Asti
+    Cuneo motorway, in connection with which the Italian authorities have indeed proposed
+    a series of measures.
+
+    As the discussions between the Commission and the Italian authorities are ongoing,
+    the Commission is not in a position to comment on the details of those discussions.'
+- source_sentence: 'query: European Commission ‘guidelines for inclusive communication’
+
+    In its ''guidelines for inclusive communication’, the European Commission once
+    again demonstrates its firm rejection of European traditions and cultural heritage,
+    telling people, for example, to avoid using the word ‘Christmas’.
+
+    1. Which institution drew up this document?
+
+    2. How much was spent on it?
+
+    3. Were external people, companies or other bodies such as NGOs involved in its
+    making and if so, which ones?'
+  sentences:
+  - 'passage: The EU supports the Tourism and Cultural Heritage sectors in Egypt[1].
+    In December 2017, in the framework of a call for proposals[2], the project "Revival
+    of the Shali Fortress" was awarded a grant of EUR 539,078.96, constituting up
+    to 90% of the total budget of the action. The premise of the initiative is that
+    support to cultural heritage stimulates socio-economic development of the communities
+    around the sites. The EU is supporting both tangible and intangible cultural heritage
+    as factors for the expansion of sustainable tourism in an economically depressed
+    area of Egypt. This project expects to result in the upgrade of 30 houses of local
+    people working in the tourist sector, open a museum/cultural centre, open a Mother
+    and Child Health Centre, and ensure that at least 50% of women in the community
+    visit the Health Centre by the end of 2019. Moreover, the project aims to ensure
+    the permanent employment of 30 trained men and youth in the tourism sector, and
+    70 enterprises will be financially empowered through a loan portfolio of approximately
+    EGP 1 million.
+
+    [1] C(2015)5244 final of 22.7.2015.
+
+    [2] reference EuropeAid/154927/DD/ACT/EG Cultural Heritage for Social and Economic
+    Development'
+  - 'passage: The Commission is fully committed to a Union of Equality and respect
+    for all. The Preamble of the Treaty on European Union refers to the ‘inspiration
+    drawn from the cultural, religious and humanist inheritance of Europe, from which
+    have developed the universal human rights, freedom, democracy, equality and the
+    rule of law’.
+
+    The guidelines were an internal document intended only for the use of staff. Its
+    aim was to sensitise Commission officials to the diversity of European culture
+    and people and to promote an inclusive attitude toward all walks of life and belief
+    of European citizens. The document has been withdrawn.
+
+    The internal document to which the Honourable Member refers did not preclude the
+    use of any words or names, like Christmas, and imposed no obligations on Commission
+    staff members. Celebrating Christmas, as well as using Christian names, are part
+    of the rich European heritage.'
+  - 'passage: The internal guidance document, to which the Honourable Member refers,
+    was an internal document intended only for the use of Commission staff. Its aim
+    was to sensitise them to the diversity of European culture and people and to promote
+    inclusion towards all walks of life and the diverse beliefs of European citizens.
+    The document has since been withdrawn.
+
+    The internal guidelines were prepared by Commission staff as part of their regular
+    tasks.'
+- source_sentence: 'query: EU funding in Rhineland-Palatinate 2014-2020
+
+    1. By the end of 2018, how much funding had Rhineland-Palatinate received from
+    the European Regional Development Fund (ERDF), the European Agricultural Guarantee
+    Fund (EAGF), the European Agricultural Fund for Rural Development (EAFRD), the
+    European Social Fund (ESF), the Lifelong Learning Programme (LLP) and other EU
+    programmes?
+
+    2. What proportion of this funding was co-financed by the Federal Republic of
+    Germany or by the federal state of Rhineland-Palatinate?
+
+    3. What is the total amount of funding still available to Rhineland-Palatinate
+    under these programmes for the 2014-2020 funding period?'
+  sentences:
+  - 'passage: The EU is aware of the case raised by the Honourable Member but is not
+    aware of any recent development in this respect.
+
+    Fully aware of the human cost suffered as a result of the Western Sahara conflict,
+    the EU supports the renewed United Nations (UN) efforts towards a solution of
+    this protracted dispute, including the most recent resolution of the UN Security
+    Council (2468)[1]. Meanwhile, should the parties so decide and under the aegis
+    of the UN, the EU would be ready to consider supporting confidence building measures,
+    including initiatives facilitating family reunification.
+
+    [1] http://unscr.com/en/resolutions/2468'
+  - 'passage: The Commission confirms that the project ‘M.Punkt RLP - Mach’ Dein Dorf’
+    was co-financed from the European Social Fund (ESF) programme Rhineland-Palatinate
+    between 2009 and 2013.
+
+    1. The ESF in Germany is implemented in one national and several regional programmes
+    (e.g. Rhineland-Palatinate) and it is managed under shared management rules. Member
+    States are responsibile for selecting the projects (e.g. assessing the proportionality
+    of a given project, trustworthiness of the beneficiaries) and controlling and
+    auditing them.
+
+    2. The ESF support for the project ‘M.Punkt RLP - Mach’ Dein Dorf’ ended in 2013.
+    Since 2014 the project has been receiving national funding from the Ministry of
+    the Interior and Sport of Rhineland-Palatinate.
+
+    3. The Commission measures the impact of an ESF programme by using indicators
+    (output/result), without looking specifically at project level. However, to measure
+    the impact of a given ESF project (e.g. labour market integration) an assessment
+    is carried out post-implementation. Such an assessment is within the scope of
+    authority of the Member State based on the shared management principle.'
+  - 'passage: The Honourable Member can find readily available information in response
+    to its question in Annex to this reply.'
+- source_sentence: 'query: Most widely spoken mother tongues in the EU
+
+    The last survey on the most widely spoken mother tongues in the EU was the one
+    by Eurobarometer in 2012, before Brexit. Eurostat does not have this information
+    either.
+
+    Bearing in mind the great change in the EU’s population brought about by the withdrawal
+    of the United Kingdom, with the loss of over 60 million people, up-to-date information
+    on this is now needed.
+
+    Is the Commission planning to commission a study on mother tongues from Eurostat
+    or to include a section on this in the next Eurobarometer?'
+  sentences:
+  - 'passage: 1. The European Border and Coast Guard Agency (Frontex) is providing
+    substantial support to notably Lithuania and Latvia. The Commission has asked
+    Frontex to provide a response to the question raised by the Honourable Member
+    regarding the Agency’s assistance in protecting the external borders of the Union
+    with Belarus. The Agency’s reply will be sent by the Commission to the Honourable
+    Member as soon as possible.
+
+    2. The Commission considers that Belarus is engaging in a hybrid attack intended
+    to destabilise Europe as set out by the Commission President in the 2021 State
+    of the Union address.'
+  - 'passage: The effect of Brexit on the total percentage of native English speakers
+    in the EU can be derived from the 2012 Eurobarometer data. The numbers of native
+    speakers of other official languages has not significantly changed in absolute
+    terms since then. However, given that the EU27 population size (447.7 million)
+    is smaller than pre-Brexit EU (513.5 million), the relative proportions of non-English
+    mother tongue speakers have all increased accordingly.
+
+    At this stage, the Commission is not planning any further study on mother tongues
+    from Eurostat or within the next Eurobarometer.'
+  - 'passage: The Commission does not have sufficient information to answer the question.
+
+    In 2012 the European Commission conducted the Special Eurobarometer survey number
+    386, ‘Europeans and their languages’.[1] Question 48a of that survey invited respondents
+    to specify which language they considered to be their mother tongue.
+
+    Caution should however be used in interpreting those data, given the extremely
+    small sample involved. Moreover the notion of ‘native’ or ‘mother tongue’ speaker
+    is not exempt from ambiguity: it can refer to the first language spoken, the language
+    one identifies with or is identified with by others, the language he or she prefers
+    to speak etc. Several languages could be indicated as mother tongue by each respondent
+    in the Eurobarometer survey.
+
+    [1] See the Report (http://ec.europa.eu/public_opinion/archives/ebs/ebs_386_en.pdf)
+    and its Annex (http://ec.europa.eu/public_opinion/archives/ebs/ebs_386_anx_en.pdf)
+    .The entire dataset can be consulted on the European Union Open Data Portal: https://open-data.europa.eu/en/data/dataset/S1049_77_1_EBS386/resource/fb4a8cb6-5e3b-4867-9780-0b7b3ab80866'
+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: Most widely spoken mother tongues in the EU\nThe last survey on the most widely spoken mother tongues in the EU was the one by Eurobarometer in 2012, before Brexit. Eurostat does not have this information either.\nBearing in mind the great change in the EU’s population brought about by the withdrawal of the United Kingdom, with the loss of over 60 million people, up-to-date information on this is now needed.\nIs the Commission planning to commission a study on mother tongues from Eurostat or to include a section on this in the next Eurobarometer?',
+    'passage: The effect of Brexit on the total percentage of native English speakers in the EU can be derived from the 2012 Eurobarometer data. The numbers of native speakers of other official languages has not significantly changed in absolute terms since then. However, given that the EU27 population size (447.7 million) is smaller than pre-Brexit EU (513.5 million), the relative proportions of non-English mother tongue speakers have all increased accordingly.\nAt this stage, the Commission is not planning any further study on mother tongues from Eurostat or within the next Eurobarometer.',
+    'passage: The Commission does not have sufficient information to answer the question.\nIn 2012 the European Commission conducted the Special Eurobarometer survey number 386, ‘Europeans and their languages’.[1] Question 48a of that survey invited respondents to specify which language they considered to be their mother tongue.\nCaution should however be used in interpreting those data, given the extremely small sample involved. Moreover the notion of ‘native’ or ‘mother tongue’ speaker is not exempt from ambiguity: it can refer to the first language spoken, the language one identifies with or is identified with by others, the language he or she prefers to speak etc. Several languages could be indicated as mother tongue by each respondent in the Eurobarometer survey.\n[1] See the Report (http://ec.europa.eu/public_opinion/archives/ebs/ebs_386_en.pdf) and its Annex (http://ec.europa.eu/public_opinion/archives/ebs/ebs_386_anx_en.pdf) .The entire dataset can be consulted on the European Union Open Data Portal: https://open-data.europa.eu/en/data/dataset/S1049_77_1_EBS386/resource/fb4a8cb6-5e3b-4867-9780-0b7b3ab80866',
+]
+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, 0.7042, 0.6063],
+#         [0.7042, 1.0000, 0.7004],
+#         [0.6063, 0.7004, 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: 16,873 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: 83 tokens</li><li>mean: 258.83 tokens</li><li>max: 512 tokens</li></ul> | <ul><li>min: 89 tokens</li><li>mean: 380.1 tokens</li><li>max: 512 tokens</li></ul> | <ul><li>min: 87 tokens</li><li>mean: 447.16 tokens</li><li>max: 512 tokens</li></ul> |
+* Samples:
+  | anchor                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                     | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                          |
+  |:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+  | <code>query: Afghanistan: management of European funds after the Taliban takeover<br>Following the G7 videoconference on 24 August 2021, the Commission announced that humanitarian aid to Afghanistan would be almost quadrupled.<br>Ursula von der Leyen said that the Commission would now disburse EUR 200 million in humanitarian aid to Afghanistan for 2021, rather than EUR 57 million as initially planned.<br>European financial aid to Afghanistan for the period 2021-2027 amounts to EUR 1 billion.<br>The move came as even more of a surprise given that the European institutions have just urgently repatriated EU personnel and evacuated local support staff, and that all international organisations and Western governments - in particular European ones - have closed their local offices.<br>Despite this, Commission spokesman Balazs Ujvari said there would be no interruption in the EU’s humanitarian engagement in Afghanistan.<br>In order to minimise the risk of humanitarian aid being misappropriated by the Taliban, the ar...</code> | <code>passage: The humanitarian donor community, including the EU, is committed to staying and delivering urgent life-saving assistance to the Afghan people, in full respect of the humanitarian principles of humanity, neutrality, impartiality and independence, as enshrined in the Treaties and in the European Consensus on Humanitarian Aid[1]. Therefore, there will be no interruption in the EU’s humanitarian engagement in Afghanistan.<br>The EU does not offer any support to the Taliban and will not channel its assistance to the population via structures or systems controlled by the Taliban. Only humanitarian aid is currently provided through the EU’s established partners on the ground (United Nations agencies and non-governmental organisations), fully in line with the international humanitarian principles mentioned above.<br>The Commission President announced on 12 October 2021 an Afghan Support Package amounting to around EUR 1 billion, which will also include ‘humanitarian+’[2] funding in direct s...</code> | <code>passage: The Commission has substantially increased EU humanitarian aid for Afghans since the Taliban takeover of Kabul on 15 August 2021 to EUR 227 million[1] until 29 November 2021 since the start of the year. These funds will address urgent humanitarian needs mainly within Afghanistan, with a particular focus on food security, nutrition, education, winterisation, protection, water, sanitation and hygiene[2] and logistics. All the funds were contracted by 10 December 2021. A smaller share (EUR 5 million) has been contracted to support Afghan refugees in neighbouring countries.<br>The humanitarian community, including the EU, is committed to staying and delivering life-saving assistance to vulnerable Afghans. This requires that independence, security and unhindered access for humanitarian staff - including all female humanitarian workers - are allowed and facilitated in full respect of the humanitarian principles of humanity, neutrality, impartiality and independence.<br>In addition, and in...</code>    |
+  | <code>query: Declaration of interests of Commissioner Stella Kyriakides<br>In her declaration of interests, European Commissioner for Health Stella Kyriakides listed a number of management positions held by her husband, Kyriakos Kyriakides, but not his management position at Maralo Limited. According to information published by the Audit Office of the Republic of Cyprus, Kyriakos Kyriakides is the managing director of Maralo Limited and the company has received several million in loans. Given that Commissioner Stella Kyriakides is playing an important role in concluding the EU’s contracts with COVID-19 vaccine manufacturers and for the purposes of transparency in that process, I should like to ask:<br>1. Can the Commission confirm whether Kyriakos Kyriakides, the husband of Commissioner Stella Kyriakides, holds a management position at Maralo Limited and, if so, how long has he held that position?<br>2. Why was this position not mentioned in the Commissioner’s declaration of interests?<br>3. If the disp...</code>       | <code>passage: 1. Based on the information received from Commissioner Kyriakides, the Commission can confirm that her husband, Mr Kyriakos Kyriakides, has been a Director of Maralo Ltd., a company in the housing and construction sector, since November 2006.<br>2. The Commissioner has explained that the company has ceased its business activities since many years and has limited itself to the liquidation of its assets. This was the reason for not declaring the position, contrary to her husband’s other positions held as declared.<br>The Code of Conduct for the Members of the Commission requires declaring financial interests of a spouse where those might be capable of giving rise to a conflict of interest. In the present case, the Cyprus Cooperative Bank, which was a large bank in Cyprus, granted commercial loans to Maralo Ltd before the year 2013. Given the commercial character of the loan and of the bank, and the company’s field of activity, a conflict of interest as defined in the Code[1] is not fo...</code> | <code>passage: Commissioner Kyriakides has declared in her declaration of interest to own a 13% share of the company ‘A. Triantafyllides Estates Ltd’.<br>She has informed the President that ‘A. Triantafyllides Estates Ltd’ is a family company whose shares originally belonged to her father and his two brothers. The letter “A” stands for Antonis, the name of the Commissioner’s grandfather. This company erected a building on the site where the house of the Commissioner’s grandparents was located and subsequently sold it. The company is now and has been for over 30 years a company without any activities whatsoever and derives no income. Its shareholders[1] have no financial benefits resulting of their shareholding.<br>Commissioner Kyriakides also informed the President that ‘Antis Triantafyllides & Sons LLC’ is a law firm which has no professional connection with the company ‘A. Triantafyllides Estates Ltd’ nor the opposite. The activities of the law firm in no way affect the company or the other way ...</code>    |
+  | <code>query: Covid-19<br>The Covid-19 crisis is taking on enormous proportions, not only in terms of the number of cases and deaths, but also as regards the measures that Member States are taking to slow its spread.<br>Every day we are seeing data and statistics comparing the development of the disease in different countries, but it is clear that there is no common standard for collecting that data. The different countries have adopted diagnostic protocols in line with their capabilities, but if we want to compare data in order to build up an accurate picture of the development of the pandemic at European level, we need uniform data, or we need to come up with indirect measurement arrangements or estimates that allow for standardised comparisons to be made.<br>It seems clear that the data on mortality or ICU admissions are the most reliable because they come from confirmed positives, and even though there are differences in the data collected, they do not seem to be as uniform as the data on incidenc...</code>          | <code>passage: The Commission is working with Member States and international partners on all fronts to tackle the COVID-19 outbreak[1]. The first priority is to guarantee the health and safety of everyone: protecting people from the spread of the virus, supporting the health systems and health workers while maintaining the flow of goods[2], mitigating the effects on the economy[3] and helping people get back to their homes.<br>Timely and accurate laboratory testing from cases under investigation is an essential part of the clinical management of COVID-19. Testing is also a key component to understanding the epidemiological situation and will be key to the phasing out of confinement measures. Recommendations on testing for COVID-19 are aligned between the European Centre for Disease Prevention and Control (ECDC)[4] and the World Health Organization[5]. To support Member States in effectively using testing tools in the context of their national strategies and during the different stages of the p...</code>    | <code>passage: The Commission is working with Member States and international partners on all fronts to tackle the COVID-19 outbreak and it has set up a website to keep all citizens updated about the evolution of the disease in the EU and informed of the comprehensive response by the Commission[1].<br>Data collection is a crucial part of monitoring the epidemiological progress of the virus. The European Centre for Disease Prevention and Control follows closely and publishes daily statistical information about confirmed cases of COVID-19[2] and has provided guidance to Member States on the collection of data, including on statistics related to the number of corona related deaths.<br>Eurostat collects[3] data on health care resources and facilities: availability and occupancy of hospital beds, technical resources and medical technology, health care personnel and health care expenditure[4].<br>Further, responding to the increased need for statistical information, Eurostat launched in April 2020 a new da...</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`: 13
+- `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`: 13
+- `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 |
+|:------:|:----:|:-------------:|
+| 0.0947 | 50   | 1.5266        |
+| 0.1894 | 100  | 0.4477        |
+| 0.2841 | 150  | 0.3848        |
+| 0.3788 | 200  | 0.3411        |
+| 0.4735 | 250  | 0.3158        |
+| 0.5682 | 300  | 0.3064        |
+| 0.6629 | 350  | 0.2887        |
+| 0.7576 | 400  | 0.2796        |
+| 0.8523 | 450  | 0.2516        |
+| 0.9470 | 500  | 0.2673        |
+
+
+### Training Time
+- **Training**: 4.8 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.*
+-->

v1.0/experiments/models/me5-base-ft2019/config.json

@@ -0,0 +1,30 @@
+{
+  "add_cross_attention": false,
+  "architectures": [
+    "XLMRobertaModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "is_decoder": false,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "xlm-roberta",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "output_past": true,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "tie_word_embeddings": true,
+  "transformers_version": "5.20.2",
+  "type_vocab_size": 1,
+  "use_cache": false,
+  "vocab_size": 250002
+}

v1.0/experiments/models/me5-base-ft2019/config_sentence_transformers.json

@@ -0,0 +1,14 @@
+{
+  "__version__": {
+    "pytorch": "2.14.1+cu130",
+    "sentence_transformers": "6.1.0",
+    "transformers": "5.20.2"
+  },
+  "default_prompt_name": null,
+  "model_type": "SentenceTransformer",
+  "prompts": {
+    "document": "",
+    "query": ""
+  },
+  "similarity_fn_name": "cosine"
+}

v1.0/experiments/models/me5-base-ft2019/modules.json

@@ -0,0 +1,20 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.base.modules.transformer.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.sentence_transformer.modules.pooling.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.base.modules.normalize.Normalize"
+  }
+]

v1.0/experiments/models/me5-base-ft2019/parlir_train_meta.json

@@ -0,0 +1,12 @@
+{
+ "base": "intfloat/multilingual-e5-base",
+ "n_pairs": 16873,
+ "epochs": 1,
+ "batch_size": 32,
+ "lr": 2e-05,
+ "max_len": 512,
+ "seed": 13,
+ "hard_negatives": true,
+ "min_year": 2019,
+ "build": "data/build/v1.0"
+}

v1.0/experiments/models/me5-base-ft2019/sentence_bert_config.json

@@ -0,0 +1,10 @@
+{
+    "transformer_task": "feature-extraction",
+    "modality_config": {
+        "text": {
+            "method": "forward",
+            "method_output_name": "last_hidden_state"
+        }
+    },
+    "module_output_name": "token_embeddings"
+}

v1.0/experiments/models/me5-base-ft2019/tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "add_prefix_space": true,
+  "backend": "tokenizers",
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 512,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "tokenizer_class": "XLMRobertaTokenizer",
+  "unk_token": "<unk>"
+}

v1.0/experiments/models/me5-base-ftdm/1_Pooling/config.json

@@ -0,0 +1,5 @@
+{
+    "embedding_dimension": 768,
+    "pooling_mode": "mean",
+    "include_prompt": true
+}

v1.0/experiments/models/me5-base-ftdm/2_Normalize/config.json

@@ -0,0 +1,4 @@
+{
+    "module_input_name": "sentence_embedding",
+    "module_output_name": "sentence_embedding"
+}

v1.0/experiments/models/me5-base-ftdm/README.md

@@ -0,0 +1,526 @@
+---
+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: Spanish failure to submit monthly reports regarding Gibraltar
+
+    What action does the Commission propose to take against Spain following its failure
+    to submit monthly reports describing the risks, actions and results of the situation
+    at its land frontier with Gibraltar, as requested by the Commission in its letter
+    of 30 July 2014?'
+  sentences:
+  - 'passage: Upon request from the Commission, the Spanish authorities have submitted
+    on 20 February the reports on the checks carried out as well as the number of
+    seizure, administrative offences or crimes related to tobacco smuggling. These
+    reports cover the period May - December 2014.
+
+    The Commission is continuing the dialogue with Spanish authorities in order to
+    ensure that the relevant data and reports are being timely submitted.'
+  - 'passage: The Spanish authorities have submitted the Action protocol establishing
+    guidelines on risk analysis at its land border with Gibraltar on 2 February 2015.
+    The monthly reports to which the Honourable Member is referring have not been
+    provided by Spain so far.'
+  - 'passage: The general organisation of the Presidency of the Council of the EU
+    is a matter for the Member State concerned and the General Secretariat of the
+    Council.
+
+    The Commission and its services cooperate closely with all Member States before
+    and during their term in the Presidency of the Council of the EU.'
+- source_sentence: 'query: UK requests for assistance from the EU Food Fraud Network
+
+    On how many occasions in the last year have the UK Food Standards Agency and Food
+    Standards Scotland requested assistance through the EU Food Fraud Network, and
+    in relation to what specific matters?'
+  sentences:
+  - 'passage: In 2017, the UK Food Standards Agency and Food Standards Scotland requested
+    assistance through the EU Food Fraud Network (FFN) on 10 occasions. The subjects
+    of these requests were mislabelling, online sale of substances unfit for human
+    consumption, absent/falsified documentation, unapproved treatment with pesticides,
+    organic certification fraud, product/ingredient substitution and suspected illegal
+    harvesting.'
+  - 'passage: The Council preparatory bodies are currently examining the proposal
+    on air passenger rights. The Council is not in a position to anticipate the outcome
+    and the duration of the ongoing discussions.'
+  - 'passage: In 2018, the United Kingdom Food Standards Agency and Food Standards
+    Scotland requested assistance through the EU Food Fraud Network on seven occasions.
+    These requests related to suspicions of mislabelling, addition of undeclared/unauthorized
+    substances, nutrition/health claims, counterfeit goods (trademarks), product replacement/substitution
+    and online sale of substances unfit for human consumption.'
+- source_sentence: 'query: Chief Scientific Adviser role axed
+
+    In light of the recent decision to axe the role of Chief Scientific Adviser, how
+    will the European Political Strategy Centre ensure that decisions are supported
+    by independent scientific evidence?'
+  sentences:
+  - 'passage: The Commission would refer the Honourable Member to its answer to written
+    questions
+
+    [REF] by M. Hans-Olaf Henkel, [REF] by M. Marc Tarabella and
+
+    [REF] by Ms. Neena Gill.'
+  - 'passage: The European Commission has not taken any decision to "abolish" the
+    position of Chief Scientific Adviser. The mandate of the Chief Scientific Adviser
+    was linked to the mandate of the previous President of the Commission and expired
+    with the end of his mandate on 31 October 2014.
+
+    The European Commission listens and consults a multitude of actors and solicits
+    a variety of views, which are subsequently scrutinised by a very thorough impact
+    assessment process, before tabling proposals. It has, therefore, a proven record
+    of rigorous, transparent and evidence based decision making.
+
+    Commissioner for Research, Science and Innovation has been asked to reflect and
+    present options on how to better institutionalise future independent scientific
+    advice to the Commission, based on the experience made in all Member States.'
+  - 'passage: The Commission is not presently in a position to undertake the research
+    that a detailed answer to the Honourable Member’s question would require. The
+    decentralised nature of the advertising spending in the Commission, as well as
+    the developing significant impact of the Covid pandemic on previously planned
+    or ongoing communication campaigns would render the process subject to revision
+    throughout the year.'
+- source_sentence: 'query: Information on possible EU sanctions in respect of ATP
+    S.p.A.
+
+    Will the Commission provide information to establish whether the company ATP Azienda
+    Trasporti Provinciali S.p.A. (registered office Via Conturli 53-16042, Carasco
+    (GE)) has ever received EU funding intended to modernise the company’s vehicle
+    fleet and in particular equip each bus with a GPS device?
+
+    If this is the case, will the Commission also say whether it has initiated infringement
+    proceedings against Italy because of the failure or partial failure of the company
+    in question to implement the projects which had attracted EU co-funding?'
+  sentences:
+  - 'passage: No, the Commission has not carried out an assessment of the eventuality
+    referred to by the Honourable Member.'
+  - 'passage: According to the information provided by the managing authority, the
+    project mentioned by the Honorable Member has not been co-financed by the European
+    Regional Development Fund.'
+  - 'passage: The Commission would like to thank the Honourable Member for the information
+    provided in his question about access to the activity of ATP experts in Italy.
+    However, the Commission does not possess sufficiently detailed information on
+    the legislation in force in Italy concerning this activity in order to judge whether
+    the limitations that the Honourable Member refers to are duly justified, proportionate
+    and necessary in the light of Directive 2006/123/EC on services in the internal
+    market ("the Services Directive").
+
+    The Commission will contact the Italian authorities to obtain more information.
+    The Commission will inform the Honourable Member of its assessment of the applicable
+    Italian legislation.'
+- source_sentence: 'query: Coding for the sharing economy
+
+    In its answer to Written Question [REF], the Commission was unable to answer the
+    question regarding the amount of money spent in the sharing economy by officials
+    on mission because sufficient coding was not available. Beyond the ability to
+    answer this question, does the Commission believe that it may be useful to code
+    for such expenses?'
+  sentences:
+  - 'passage: The Commission will kindly refer the Honourable Member to written answer
+    [REF][1].
+
+    [1] http://www.europarl.europa.eu/plenary/en/parliamentary-questions.html'
+  - 'passage: The Commission regrets to inform the Honourable Member that it does
+    not have sufficient information to be able to answer the question of the amount
+    of money spent in the sharing economy by officials on mission. The IT systems
+    used to manage and reimburse expenses incurred whilst on official business focuses
+    on amount, type and proof of the expenses and does not provide for encoding of
+    precise references to the companies concerned. However, Commission Staff are allowed
+    to use sharing economy services, if they can provide an invoice, which is the
+    case for an increasing number of those companies.'
+  - 'passage: The Commission and its staff are bound to comply with the rules laid
+    down by the Financial Regulation, the Staff Regulations and its implementing rules,
+    such as the Guide to missions. Therefore, the Commission''s staff arrange their
+    trips according to the rule of cost-effectiveness. In this context, whenever the
+    requirements of the Guide to missions are met, the use of sharing economy services
+    is permitted.
+
+    The possibility to enable coding for such expenses in the IT tool could be envisaged
+    in the context of broader legal and budgetary feasibility analysis.'
+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: Coding for the sharing economy\nIn its answer to Written Question [REF], the Commission was unable to answer the question regarding the amount of money spent in the sharing economy by officials on mission because sufficient coding was not available. Beyond the ability to answer this question, does the Commission believe that it may be useful to code for such expenses?',
+    "passage: The Commission and its staff are bound to comply with the rules laid down by the Financial Regulation, the Staff Regulations and its implementing rules, such as the Guide to missions. Therefore, the Commission's staff arrange their trips according to the rule of cost-effectiveness. In this context, whenever the requirements of the Guide to missions are met, the use of sharing economy services is permitted.\nThe possibility to enable coding for such expenses in the IT tool could be envisaged in the context of broader legal and budgetary feasibility analysis.",
+    'passage: The Commission regrets to inform the Honourable Member that it does not have sufficient information to be able to answer the question of the amount of money spent in the sharing economy by officials on mission. The IT systems used to manage and reimburse expenses incurred whilst on official business focuses on amount, type and proof of the expenses and does not provide for encoding of precise references to the companies concerned. However, Commission Staff are allowed to use sharing economy services, if they can provide an invoice, which is the case for an increasing number of those companies.',
+]
+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, 0.7865, 0.6481],
+#         [0.7865, 1.0000, 0.5836],
+#         [0.6481, 0.5836, 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: 47 tokens</li><li>mean: 225.93 tokens</li><li>max: 403 tokens</li></ul> | <ul><li>min: 53 tokens</li><li>mean: 320.8 tokens</li><li>max: 512 tokens</li></ul> | <ul><li>min: 147 tokens</li><li>mean: 387.31 tokens</li><li>max: 512 tokens</li></ul> |
+* Samples:
+  | anchor                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               | positive                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       | negative                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                          |
+  |:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+  | <code>query: Co-financing of projects<br>The European rules of governance and the pressure of debt markets in which the ECB has refused to intervene have caused many Member States to make drastic adjustments to their budgets. As a result of the rules on co-financing, the repercussions of these cuts snowball when they prevent access to EU funds in projects of mutual interest.<br>Given the above, will the Commission reduce the co-financing rates for projects by Member States' public administrations, particularly for social cohesion and convergence projects, such as those financed under the European Regional Development Fund (ERDF) or the European Social Fund (ESF)?<br>Will the Commission extend deadlines so that Member States can pay their contributions, thus postponing the payments of Member States with serious budget difficulties?<br>Finally, given that the Commission is creating new initiatives - such as the European Fund for Strategic Investments (EFSI) - which are aimed exclusively at the mobilisatio...</code> | <code>passage: 1. Regulation (EU) N° 1297/2013 of the European Parliament and the Council of 11 December 2013[1] allows payments under the 2007-2013 programmes to be increased by an amount corresponding to 10 % above the co-financing rate applicable to each priority, but not exceeding 100%, where Member States benefit from one of the EU financial assistance mechanisms referred to in the regulation. This increase applies to the amount of eligible expenditure newly declared in each certified statement of expenditure submitted until the end of the programming period.<br>2. The final date of eligibility of expenditure is 31 December 2015 under the 2007-2013 period and this deadline will not be extended. Eligible expenditure means expenditure incurred and paid by the beneficiary. Council Regulation 1083/2006 of 11 July 2006[2] does not specify any final date for the reimbursement to the beneficiaries of the public contribution but states that it should be paid "as quickly as possible".<br>3. The Commis...</code> | <code>passage: ([d.m.yyyy])<br>The Integrated Mediterranean Programmes were established for seven years (1986-1992), in order to help ‘the southern regions of the Community’ - defined as the whole of Greece, parts of southern France and most of southern Italy - ‘to adjust under the best conditions possible to the new situation created by enlargement’. They were not renewed, but Greece, Portugal and Spain were able to benefit from the new Cohesion Fund from 1993 onwards.<br>Other measures have been taken to support European regions most affected by the crisis such as increasing the co-financing rates of many Structural Funds and Cohesion Funds programmes up to the maximum allowed by the regulation[1] which substantially reduced national co-financing. Under the 2014-2020 period, more support from the Structural Funds (the European Regional Development Fund and the European Social Fund) is allocated to less developed regions, including southern Italy, part of Greece and Croatia and the Cohesion Fund ...</code>    |
+  | <code>query: EU-Ukraine Association Agreement<br>Does the Commission accept that, given the current situation in Ukraine, the signing of the EU-Ukraine Association Agreement, which aims to ‘deepen political association’ (Article 4(2)(a)), might actually further exacerbate political tensions within that country?</code>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      | <code>passage: The EU has deeply engaged in seeking a peaceful settlement of the crisis in Ukraine. We have to find ways to explain to Russia that respect can be regained if we work jointly on a vision of cooperation on the European continent without dividing lines. The Association Agreement is not meant to replace - or intrude in - Ukraine's relations with any neighbour, but we have taken note of the public support for closer association and integration with Europe. Throughout the year, the European Union and Ukraine have engaged with Russia in order to discuss the economic effects on the Russian economy of the implementation of the AA/DCFTA with a view to clarify concerns and - where appropriate - identify solutions.<br>A concrete result of these consultations is the agreement reached at the trilateral ministerial meeting on 12 September between the EU, Ukraine and Russia. Consequently, the Commission has proposed to the Council to delay until 31 December 2015 the provisional application of t...</code>    | <code>passage: The EU notes evidence of continued support given to the separatists by Russia, which underlines Russia's responsibility in destabilizing the situation in eastern Ukraine.<br>Ukraine is currently engaged in a broad and complex reform process aiming to respond to its citizens's aspirations for a democratic, secure and prosperous country.<br>The European Union has put in place an unprecedented package of measures in support of Ukraine's reform steps, including political support, financial, technical and economic assistance, and strategic advice. The EU maintains close coordination with the Ukrainian authorities to ensure the commitments undertaken within the Association Agreement framework are fully implemented. The EU-Ukraine Association Agenda is in this context the main political tool for the implementation and monitoring of the Association Agreement.<br>The EU has also joined the international efforts and stepped-up its humanitarian assistance to the people affected by the conflict in...</code> |
+  | <code>query: CBRN preparedness and protection measures<br>On 17 January 2018, the World Economic Forum published on its latest annual Global Risks Report. The likelihood that weapons of mass destruction might be used has again increased compared with last year’s assessment. International reports provide more and more reasons to believe that the threat of the deliberate use of CBRN weapons is becoming more plausible. In October 2017, the Commission adopted an Action Plan to enhance preparedness against chemical, biological, radiological and nuclear security risks, containing several proposals aimed at supporting Member States in better defending EU citizens against terrorist threats. However, the level of risk awareness and, therefore, preparedness in the event of a CBRN incident differs widely between all 28 Member States.<br>1. Knowing that only a few Member States would be sufficiently protected in the event of a CBRN attack, and knowing that such attacks typically have a cross-border impact, wh...</code>       | <code>passage: Protection of citizens and infrastructures remains primarily responsibility of Member States of the European Union. Nevertheless in the Action Plan to enhance preparedness against chemical, biological, radiological and nuclear security risks[1] adopted on 18 October 2017, the Commission proposed several initiatives focusing on sharing of expertise and resources among Member States aiming at better preparedness of the EU as a whole for the chemical, biological, radiological and nuclear (CBRN) terrorist threats. Among others the Commission proposed development of a common EU CBRN training curriculum, boosting of cross-border trainings and exercises and encouraging Member States to commit more capacities to the European Emergency Response Capacity (EERC) of the Union's Civil Protection Mechanism. These - and other initiatives announced in the Action Plan - should over time result in an enhanced preparedness of the EU for the CBRN terrorist threats.<br>[1] COM(2017) 610 final</code>               | <code>passage: Internal security, including the protection of citizens, residents and infrastructures is the primary responsibility of the Member States of the European Union.<br>Nevertheless, in the Action Plan to enhance preparedness against chemical, biological, radiological and nuclear (CBRN) security risks[1] adopted on 18 October 2017, the Commission proposed several initiatives focusing on sharing of expertise and resources among Member States aiming at better preparedness of the EU as a whole against threats from CBRN agents. Among others, the Commission proposed the development of a common EU CBRN training curriculum or boosting of cross-border trainings. The initiatives announced in the Action Plan will over time result in an enhanced preparedness of the EU against CBRN security risks as well as a higher security level of EU citizens and residents.<br>Moreover, following the Salisbury attack and at the invitation of the European Council[2], the Commission and the High Representative prop...</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`: 13
+- `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`: 13
+- `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   | 2.2578        |
+| 0.0537 | 100  | 0.6298        |
+| 0.0805 | 150  | 0.4273        |
+| 0.1074 | 200  | 0.4102        |
+| 0.1342 | 250  | 0.3654        |
+| 0.1610 | 300  | 0.3740        |
+| 0.1879 | 350  | 0.3531        |
+| 0.2147 | 400  | 0.3299        |
+| 0.2415 | 450  | 0.3342        |
+| 0.2684 | 500  | 0.3238        |
+| 0.2952 | 550  | 0.3122        |
+| 0.3221 | 600  | 0.2966        |
+| 0.3489 | 650  | 0.3243        |
+| 0.3757 | 700  | 0.3011        |
+| 0.4026 | 750  | 0.3047        |
+| 0.4294 | 800  | 0.3229        |
+| 0.4563 | 850  | 0.2850        |
+| 0.4831 | 900  | 0.2881        |
+| 0.5099 | 950  | 0.2753        |
+| 0.5368 | 1000 | 0.2791        |
+| 0.5636 | 1050 | 0.2995        |
+| 0.5904 | 1100 | 0.2569        |
+| 0.6173 | 1150 | 0.2742        |
+| 0.6441 | 1200 | 0.2807        |
+| 0.6710 | 1250 | 0.2950        |
+| 0.6978 | 1300 | 0.2713        |
+| 0.7246 | 1350 | 0.2697        |
+| 0.7515 | 1400 | 0.2640        |
+| 0.7783 | 1450 | 0.2753        |
+| 0.8052 | 1500 | 0.2511        |
+| 0.8320 | 1550 | 0.2549        |
+| 0.8588 | 1600 | 0.2521        |
+| 0.8857 | 1650 | 0.2665        |
+| 0.9125 | 1700 | 0.2481        |
+| 0.9393 | 1750 | 0.2616        |
+| 0.9662 | 1800 | 0.2406        |
+| 0.9930 | 1850 | 0.2848        |
+
+
+### Training Time
+- **Training**: 16.1 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.*
+-->

v1.0/experiments/models/me5-base-ftdm/config.json

@@ -0,0 +1,30 @@
+{
+  "add_cross_attention": false,
+  "architectures": [
+    "XLMRobertaModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "is_decoder": false,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "xlm-roberta",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "output_past": true,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "tie_word_embeddings": true,
+  "transformers_version": "5.20.2",
+  "type_vocab_size": 1,
+  "use_cache": false,
+  "vocab_size": 250002
+}

v1.0/experiments/models/me5-base-ftdm/config_sentence_transformers.json

@@ -0,0 +1,14 @@
+{
+  "__version__": {
+    "pytorch": "2.14.1+cu130",
+    "sentence_transformers": "6.1.0",
+    "transformers": "5.20.2"
+  },
+  "default_prompt_name": null,
+  "model_type": "SentenceTransformer",
+  "prompts": {
+    "document": "",
+    "query": ""
+  },
+  "similarity_fn_name": "cosine"
+}

v1.0/experiments/models/me5-base-ftdm/modules.json

@@ -0,0 +1,20 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.base.modules.transformer.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.sentence_transformer.modules.pooling.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.base.modules.normalize.Normalize"
+  }
+]

v1.0/experiments/models/me5-base-ftdm/parlir_train_meta.json

@@ -0,0 +1,18 @@
+{
+ "base": "intfloat/multilingual-e5-base",
+ "n_pairs": 59612,
+ "epochs": 1,
+ "batch_size": 32,
+ "lr": 2e-05,
+ "max_len": 512,
+ "seed": 13,
+ "hard_negatives": true,
+ "min_year": 0,
+ "max_year": 0,
+ "hardneg_window_days": 365,
+ "negative_stats": {
+  "date_matched_bm25": 58288,
+  "date_matched_random": 1324
+ },
+ "build": "data/build/v1.0"
+}

v1.0/experiments/models/me5-base-ftdm/sentence_bert_config.json

@@ -0,0 +1,10 @@
+{
+    "transformer_task": "feature-extraction",
+    "modality_config": {
+        "text": {
+            "method": "forward",
+            "method_output_name": "last_hidden_state"
+        }
+    },
+    "module_output_name": "token_embeddings"
+}

v1.0/experiments/models/me5-base-ftdm/tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "add_prefix_space": true,
+  "backend": "tokenizers",
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 512,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "tokenizer_class": "XLMRobertaTokenizer",
+  "unk_token": "<unk>"
+}