Instructions to use amentaphd/eu-regulation-embeddings-snowflake-m-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use amentaphd/eu-regulation-embeddings-snowflake-m-v2 with sentence-transformers:

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("amentaphd/eu-regulation-embeddings-snowflake-m-v2", trust_remote_code=True)

sentences = [
    "What are the anticipated financial effects that could arise from material risks associated with resource use and circular economy, and how might these risks impact the financial position, performance, and cash flows of an undertaking over different time frames?",
    "(a)\n\nanticipated financial effects due to material risks arising from material resource use and circular economy -related impacts and dependencies and how these risks have or could reasonably be expected to have) a material influence on the undertaking’s financial position, financial performance performance, and cash flows over the short-, medium- and long-term; and\n\n(b)\n\nanticipated financial effects due to material opportunities related to resource use and circular economy.\n\nThe disclosure shall include:\n\n(a)",
    "combination of hydrocarbons obtained as a raffinate from a sulphuric acid treating process. It consists of hydrocarbons having carbon numbers predominantly in the range of C7 through C12 and boiling in the range of approximately 90 °C to 230 °C.) 649-351-00-7 265-115-2 64742-15-0 P Naphtha (petroleum), chemically neutralised heavy; Low boiling point naphtha — unspecified (A complex combination of hydrocarbons produced by a treating process to remove acidic materials. It consists of hydrocarbons having carbon numbers predominantly in the range of C6 through C12 and boiling in the range of approximately 65 °C to 230 °C.) 649-352-00-2 265-122-0 64742-22-9 P Naphtha (petroleum), chemically neutralised light; Low boiling point naphtha —",
    "2. Member States shall require any investment firm wishing to establish a branch within the territory of another Member State or to use tied agents established in another Member State in which it has not established a branch, first to notify the competent authority of its home Member State and to provide it with the following information:\n\n(a) the Member States within the territory of which it plans to establish a branch or the Member States in which it has not established a branch but plans to use tied agents established there;\n\n(b) a programme of operations setting out, inter alia, the investment services and/or activities as well as the ancillary services to be offered;\n\n(c) where established, the organisational structure of the branch and indicating whether the branch intends to use tied agents and the identity of those tied agents;\n\n(d) where tied agents are to be used in a Member State in which an investment firm has not established a branch, a description of the intended use of the tied agent(s) and an organisational structure, including reporting lines, indicating how the agent(s) fit into the corporate structure of the investment firm;\n\n(e) the address in the host Member State from which documents may be obtained;\n\n(f) the names of those responsible for the management of the branch or of the tied agent.\n\nWhere an investment firm uses a tied agent established in a Member State outside its home Member State, such tied agent shall be assimilated to the branch, where one is established, and shall in any event be subject to the provisions of this Directive relating to branches."
]
embeddings = model.encode(sentences)

similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [4, 4]

Notebooks
Google Colab
Kaggle

eu-regulation-embeddings-snowflake-m-v2 / README.md

amentaphd

Upload fine-tuned EU regulation embeddings model

190f9ed verified over 1 year ago

preview code

Raw

History Blame Contribute Delete

36.8 kB

metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:46338
  - loss:MatryoshkaLoss
  - loss:MultipleNegativesRankingLoss
base_model: Snowflake/snowflake-arctic-embed-m-v2.0
widget:
  - source_sentence: >-
      What are the anticipated financial effects that could arise from material
      risks associated with resource use and circular economy, and how might
      these risks impact the financial position, performance, and cash flows of
      an undertaking over different time frames?
    sentences:
      - >-
        (a)


        anticipated financial effects due to material risks arising from
        material resource use and circular economy -related impacts and
        dependencies and how these risks have or could reasonably be expected to
        have) a material influence on the undertaking’s financial position,
        financial performance performance, and cash flows over the short-,
        medium- and long-term; and


        (b)


        anticipated financial effects due to material opportunities related to
        resource use and circular economy.


        The disclosure shall include:


        (a)
      - >-
        combination of hydrocarbons obtained as a raffinate from a sulphuric
        acid treating process. It consists of hydrocarbons having carbon numbers
        predominantly in the range of C7 through C12 and boiling in the range of
        approximately 90 °C to 230 °C.) 649-351-00-7 265-115-2 64742-15-0 P
        Naphtha (petroleum), chemically neutralised heavy; Low boiling point
        naphtha — unspecified (A complex combination of hydrocarbons produced by
        a treating process to remove acidic materials. It consists of
        hydrocarbons having carbon numbers predominantly in the range of C6
        through C12 and boiling in the range of approximately 65 °C to 230 °C.)
        649-352-00-2 265-122-0 64742-22-9 P Naphtha (petroleum), chemically
        neutralised light; Low boiling point naphtha —
      - >-
        2. Member States shall require any investment firm wishing to establish
        a branch within the territory of another Member State or to use tied
        agents established in another Member State in which it has not
        established a branch, first to notify the competent authority of its
        home Member State and to provide it with the following information:


        (a) the Member States within the territory of which it plans to
        establish a branch or the Member States in which it has not established
        a branch but plans to use tied agents established there;


        (b) a programme of operations setting out, inter alia, the investment
        services and/or activities as well as the ancillary services to be
        offered;


        (c) where established, the organisational structure of the branch and
        indicating whether the branch intends to use tied agents and the
        identity of those tied agents;


        (d) where tied agents are to be used in a Member State in which an
        investment firm has not established a branch, a description of the
        intended use of the tied agent(s) and an organisational structure,
        including reporting lines, indicating how the agent(s) fit into the
        corporate structure of the investment firm;


        (e) the address in the host Member State from which documents may be
        obtained;


        (f) the names of those responsible for the management of the branch or
        of the tied agent.


        Where an investment firm uses a tied agent established in a Member State
        outside its home Member State, such tied agent shall be assimilated to
        the branch, where one is established, and shall in any event be subject
        to the provisions of this Directive relating to branches.
  - source_sentence: >-
      What steps must the single point of contact take if the project promoter
      submits an incomplete application for a Strategic Project, and how does
      this affect the permit-granting process timeline?
    sentences:
      - >-
        (1)


        ‘cooling’ means the extraction of heat from an enclosed or indoor space
        (comfort application) or from a process in order to reduce the space or
        process temperature to, or maintain it at, a specified temperature (set
        point); for cooling systems, the extracted heat is rejected into and
        absorbed by the ambient air, ambient water or the ground, where the
        environment (air, ground, and water) provides a sink for the heat
        extracted and thus functions as a cold source;


        (2)
      - >-
        1. Suppliers shall provide the manufacturer with all the information and
        documentation necessary for the manufacturer to demonstrate the
        conformity of the packaging and the packaging materials with this
        Regulation, including the technical documentation referred to in Annex
        VII and required under or pursuant to Articles 5 to 11, in one or more
        languages which can be easily understood by the manufacturer. That
        information and documentation shall be provided in either paper or
        electronic form.


        2. Where appropriate, the documentation and information required under
        Union legal acts applicable to contact-sensitive packaging shall be part
        of the information and documentation to be provided to the manufacturer
        pursuant to paragraph 1.
      - >-
        6.


        No later than 45 days following the receipt of a permit-granting
        application related to a Strategic Project, the single point of contact
        concerned shall acknowledge that the application is complete or, if the
        project promoter has not sent all the information required to process an
        application, request the project promoter to submit a complete
        application without undue delay, specifying which information is
        missing. Where the application submitted is deemed to be incomplete a
        second time, the single point of contact concerned shall not request
        information in areas not covered in the first request for additional
        information and shall be entitled only to request further evidence to
        complete the identified missing information.


        The date of the acknowledgement referred to in the first subparagraph
        shall serve as the start of the permit-granting process.


        7.


        No later than one month from the date of acknowledgement referred to in
        paragraph 6 of this Article, the single point of contact concerned shall
        draw up, in close cooperation with the project promoter and other
        competent authorities concerned, a detailed schedule for the
        permit-granting process. The schedule shall be published by the project
        promoter on the website referred to in Article 8(5). The single point of
        contact concerned shall update the schedule in the event that there are
        significant changes that potentially affect the timing of the
        comprehensive decision.


        8.


        The single point of contact concerned shall notify the project promoter
        when the environmental impact assessment report referred in Article 5(1)
        of Directive 2011/92/EU is due, taking into account the organisation of
        the permit-granting process in the Member State concerned and the need
        to allow sufficient time to assess the report. The period between the
        deadline for the submission of the environmental impact assessment
        report and the actual submission of that report shall not be counted
        towards the duration of the permit-granting process referred to in
        paragraphs 1 and 2 of this Article.


        9.
  - source_sentence: >-
      What are the requirements for energy audits to be considered compliant
      with the specified paragraph, and what role do voluntary agreements play
      in this process?
    sentences:
      - >-
        8. Member States shall develop programmes to encourage enterprises that
        are not SMEs and that are not subject to paragraph 1 or 2 to undergo
        energy audits and to subsequently implement the recommendations arising
        from those audits.


        9. Energy audits shall be considered to comply with paragraph 2 where
        they are:


        (a) carried out in an independent manner, on the basis of the minimum
        criteria set out in Annex VI; (b) implemented under voluntary agreements
        concluded between organisations of stakeholders and a body appointed and
        supervised by the Member State concerned, by another body to which the
        competent authorities have delegated the responsibility concerned or by
        the Commission. --- ---
      - >-
        3.1.1. The evaluation of all available information shall comprise:


        the hazard identification based on all available information,


        the establishment of the quantitative dose (concentration)-response
        (effect) relationship.


        3.1.2. When it is not possible to establish the quantitative dose
        (concentration)-response (effect) relationship, then this should be
        justified and a semi-quantitative or qualitative analysis shall be
        included.


        3.1.3. All information used to assess the effects on a specific
        environmental sphere shall be briefly presented, if possible in the form
        of a table or tables. The relevant test results (e.g. LC50 or NOEC) and
        test conditions (e.g. test duration, route of administration) and other
        relevant information shall be presented, in internationally recognised
        units of measurement for that effect.


        3.1.4. All information used to assess the environmental fate of the
        substance shall be briefly presented, if possible in the form of a table
        or tables. The relevant test results and test conditions and other
        relevant information shall be presented, in internationally recognised
        units of measurement for that effect.


        3.1.5. If one study is available then a robust study summary should be
        prepared for that study. Where there is more than one study addressing
        the same effect, then the study or studies giving rise to the highest
        concern shall be used to draw a conclusion and a robust study summary
        shall be prepared for that study or studies and included as part of the
        technical dossier. Robust summaries will be required of all key data
        used in the hazard assessment. If the study or studies giving rise to
        the highest concern are not used, then this shall be fully justified and
        included as part of the technical dossier, not only for the study being
        used but also for all studies reaching a higher concern than the study
        being used. For substances where all available studies indicate no
        hazards an overall assessment of the validity of all studies should be
        performed.


        3.2.  Step 2 : Classification and Labelling


        ▼M51
      - >-
        impact of single-use packaging, in particular plastic carrier bags; ---
        --- (f) the composting properties and appropriate waste management
        options for compostable packaging in accordance with Article 9(2) of
        this Regulation; consumers shall be informed that compostable packaging
        is not suitable for home composting and that compostable packaging is
        not to be discarded in nature. --- ---
  - source_sentence: >-
      In what scenario should information on toxic effects be listed only once
      for a mixture?
    sentences:
      - >-
        In determining the energy savings from taxation-related policy measures
        introduced under Article 10, the following principles shall apply: (a)
        credit shall be given only for energy savings from taxation measures
        exceeding the minimum levels of taxation applicable to fuels as required
        in Council Directive 2003/96/EC (2) or 2006/112/EC (3); (b) short-run
        price elasticities for the calculation of the impact of the energy
        taxation measures shall represent the responsiveness of energy demand to
        price changes, and shall be estimated on the basis of recent and
        representative official data sources, which are applicable for the
        Member State, and, where applicable, on the basis of accompanying
        studies from an independent institute. If a different
      - >-
        Article 13


        Project development assistance


        1.


        The Commission shall, after consulting the Member States in accordance
        with Article 21(2), point (c), determine the maximum amount of
        Innovation Fund support available for project development assistance.


        2.


        The Commission may award project development assistance in the form of
        technical assistance to any project that falls within the scope of the
        Innovation Fund, as set out in Article 10a(8), first and sixth
        subparagraphs of Directive 2003/87/EC.


        3.


        The following activities may be funded by way of project development
        assistance:


        (a)


        improvement and development of project documentation or of components of
        the project design with a view to ensuring the sufficient maturity of
        the project;


        (b)


        assessment of the feasibility of the project, including technical and
        economic studies;


        (c)


        advice on the financial and legal structure of the project;


        (d)


        capacity building of the project proponent.


        4.


        If project development assistance is implemented under indirect
        management, the implementing entity shall carry out the selection
        procedure and take the decision to award the project development
        assistance after having consulted the Commission. The award criteria
        shall take into account the degree of innovation compared to the state
        of the art, the potential to significantly reduce climate impacts and to
        support widespread application, the maturity as well as the geographical
        and sectoral balance in relation to the portfolio of funded projects.
      - >-
        effects of the mixture. The information on toxic effects shall be
        presented for each substance, except for the following cases: (a)  if
        the information is duplicated, it shall be listed only once for the
        mixture overall, such as when two substances both cause vomiting and
        diarrhoea; (b)  if it is unlikely that these effects will occur at the
        concentrations present, such as when a mild irritant is diluted to below
        a certain concentration in a non-irritant solution; (c)  where
        information on interactions between substances in a mixture is not
        available, assumptions shall not be made and instead the health effects
        of each substance shall be listed separately. --- ---
  - source_sentence: >-
      How does the text suggest addressing the social aspects related to low-
      and middle-income transport users in the context of zero-emission vehicle
      initiatives?
    sentences:
      - >-
        (b)


        measures intended to accelerate the uptake of zero-emission vehicles or
        to provide financial support for the deployment of fully interoperable
        refuelling and recharging infrastructure for zero-emission vehicles, or
        measures to encourage a shift to public transport and improve
        multimodality, or to provide financial support in order to address
        social aspects concerning low- and middle-income transport users;


        (c)


        to finance their Social Climate Plan in accordance with Article 15 of
        Regulation (EU) 2023/955;


        (d)
      - >-
        If the planned change is implemented notwithstanding the first and
        second subparagraphs, or if an unplanned change has taken place pursuant
        to which the AIFM’s management of the AIF no longer complies with this
        Directive or the AIFM otherwise no longer complies with this Directive,
        the competent authorities of the Member State of reference of the AIFM
        shall take all due measures in accordance with Article 46, including, if
        necessary, the express prohibition of marketing of the AIF.
      - >-
        (d)


        for gas discharge lamps, 80 % shall be recycled.


        Part 2: Minimum targets applicable by category from 15 August 2015 until
        14 August 2018 with reference to the categories listed in Annex I:


        (a)


        for WEEE falling within category 1 or 10 of Annex I,


        85 % shall be recovered, and


        80 % shall be prepared for re-use and recycled;


        (b)


        for WEEE falling within category 3 or 4 of Annex I,


        80 % shall be recovered, and


        70 % shall be prepared for re-use and recycled;


        (c)


        for WEEE falling within category 2, 5, 6, 7, 8 or 9 of Annex I,


        75 % shall be recovered, and


        55 % shall be prepared for re-use and recycled;


        (d)


        for gas discharge lamps, 80 % shall be recycled.
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy@1
  - cosine_accuracy@3
  - cosine_accuracy@5
  - cosine_accuracy@10
  - cosine_precision@1
  - cosine_precision@3
  - cosine_precision@5
  - cosine_precision@10
  - cosine_recall@1
  - cosine_recall@3
  - cosine_recall@5
  - cosine_recall@10
  - cosine_ndcg@10
  - cosine_mrr@10
  - cosine_map@100
model-index:
  - name: SentenceTransformer based on Snowflake/snowflake-arctic-embed-m-v2.0
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: cosine_accuracy@1
            value: 0.7058518902123252
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.9067840497151735
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.9447609183497324
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9730709476954945
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.7058518902123252
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.3022613499050578
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.18895218366994648
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09730709476954946
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.7058518902123252
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.9067840497151735
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.9447609183497324
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9730709476954945
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.851314896054128
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.8109469830857718
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.8122768308333804
            name: Cosine Map@100

SentenceTransformer based on Snowflake/snowflake-arctic-embed-m-v2.0

This is a sentence-transformers model finetuned from Snowflake/snowflake-arctic-embed-m-v2.0. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: Snowflake/snowflake-arctic-embed-m-v2.0
Maximum Sequence Length: 8192 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: GteModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'How does the text suggest addressing the social aspects related to low- and middle-income transport users in the context of zero-emission vehicle initiatives?',
    '(b)\n\nmeasures intended to accelerate the uptake of zero-emission vehicles or to provide financial support for the deployment of fully interoperable refuelling and recharging infrastructure for zero-emission vehicles, or measures to encourage a shift to public transport and improve multimodality, or to provide financial support in order to address social aspects concerning low- and middle-income transport users;\n\n(c)\n\nto finance their Social Climate Plan in accordance with Article 15 of Regulation (EU) 2023/955;\n\n(d)',
    'If the planned change is implemented notwithstanding the first and second subparagraphs, or if an unplanned change has taken place pursuant to which the AIFM’s management of the AIF no longer complies with this Directive or the AIFM otherwise no longer complies with this Directive, the competent authorities of the Member State of reference of the AIFM shall take all due measures in accordance with Article 46, including, if necessary, the express prohibition of marketing of the AIF.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.7059
cosine_accuracy@3	0.9068
cosine_accuracy@5	0.9448
cosine_accuracy@10	0.9731
cosine_precision@1	0.7059
cosine_precision@3	0.3023
cosine_precision@5	0.189
cosine_precision@10	0.0973
cosine_recall@1	0.7059
cosine_recall@3	0.9068
cosine_recall@5	0.9448
cosine_recall@10	0.9731
cosine_ndcg@10	0.8513
cosine_mrr@10	0.8109
cosine_map@100	0.8123

Training Details

Training Dataset

Unnamed Dataset

Size: 46,338 training samples
Columns: sentence_0 and sentence_1
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1
type string string
details
min: 9 tokens
mean: 39.98 tokens
max: 286 tokens

min: 3 tokens
mean: 248.72 tokens
max: 1315 tokens

	sentence_0	sentence_1
type	string	string
details	min: 9 tokens mean: 39.98 tokens max: 286 tokens	min: 3 tokens mean: 248.72 tokens max: 1315 tokens

Samples:

sentence_0	sentence_1
`What is the maximum allowable reduction in excise duty for mixtures used as motor fuels containing biodiesel in Italy until 30 June 2004?`	for waste oils which are reused as fuel, either directly after recovery or following a recycling process for waste oils, and where the reuse is subject to duty. 8. ITALY: for differentiated rates of excise duty on mixtures used as motor fuels containing 5 % or 25 % of biodiesel until 30 June 2004. The reduction in excise duty may not be greater than the amount of excise duty payable on the volume of biofuels present in the products eligible for the reduction. The reduction in excise duty shall be adjusted to take account of changes in the price of raw materials to avoid overcompensating for the extra costs involved in the manufacture of biofuels;
`What are the minimum indicative share percentages for the years 2023 to 2030, and how do these percentages relate to the interconnectivity levels of the Member States?`	`Such indicative shares may, in each year, amount to at least 5 % from 2023 to 2026 and at least 10 % from 2027 to 2030, or, where lower, to the level of interconnectivity of the Member State concerned in any given year. In order to acquire further implementation experience, Member States may organise one or more pilot schemes where support is open to producers located in other Member States. 2.`
`What is the significance of the one-month period mentioned in the context?`	`one month after its notification, in accordance with the arrangements provided for in Article 23.`

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "MultipleNegativesRankingLoss",
    "matryoshka_dims": [
        768,
        512,
        256,
        128,
        64
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
num_train_epochs: 4
fp16: True
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 8
per_device_eval_batch_size: 8
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 5e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1
num_train_epochs: 4
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
save_safetensors: True
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
no_cuda: False
use_cpu: False
use_mps_device: False
seed: 42
data_seed: None
jit_mode_eval: False
use_ipex: False
bf16: False
fp16: True
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: 0
ddp_backend: None
tpu_num_cores: None
tpu_metrics_debug: False
debug: []
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_prefetch_factor: None
past_index: -1
disable_tqdm: False
remove_unused_columns: True
label_names: None
load_best_model_at_end: False
ignore_data_skip: False
fsdp: []
fsdp_min_num_params: 0
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
fsdp_transformer_layer_cls_to_wrap: None
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
dataloader_pin_memory: True
dataloader_persistent_workers: False
skip_memory_metrics: True
use_legacy_prediction_loop: False
push_to_hub: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: None
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
eval_do_concat_batches: True
fp16_backend: auto
push_to_hub_model_id: None
push_to_hub_organization: None
mp_parameters:
auto_find_batch_size: False
full_determinism: False
torchdynamo: None
ray_scope: last
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
dispatch_batches: None
split_batches: None
include_tokens_per_second: False
include_num_input_tokens_seen: False
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss	cosine_ndcg@10
0.0863	500	0.225	-
0.1726	1000	0.1337	-
0.2589	1500	0.1195	-
0.3452	2000	0.0803	-
0.4316	2500	0.0775	-
0.5179	3000	0.0714	-
0.6042	3500	0.0852	-
0.6905	4000	0.0718	-
0.7768	4500	0.0499	-
0.8631	5000	0.0665	0.8371
0.9494	5500	0.0674	-
1.0	5793	-	0.8416
1.0357	6000	0.0538	-
1.1220	6500	0.0606	-
1.2084	7000	0.0294	-
1.2947	7500	0.0129	-
1.3810	8000	0.0101	-
1.4673	8500	0.0072	-
1.5536	9000	0.0211	-
1.6399	9500	0.0133	-
1.7262	10000	0.0063	0.8513

Framework Versions

Python: 3.10.15
Sentence Transformers: 4.0.2
Transformers: 4.49.0
PyTorch: 2.6.0+cu126
Accelerate: 0.26.0
Datasets: 3.5.0
Tokenizers: 0.21.1

Citation

BibTeX

Sentence Transformers

@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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning},
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}