metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:22291
  - loss:MultipleNegativesRankingLoss
base_model: dunzhang/stella_en_1.5B_v5
widget:
  - source_sentence: >-
      What specific information should be included in the annual report to
      adequately explain a company's business model and strategy in alignment
      with Principle 7?
    sentences:
      - >
        REGULATORY REQUIREMENTS FOR AUTHORISED PERSONS ENGAGED IN REGULATED
        ACTIVITIES IN RELATION TO VIRTUAL ASSETS

        Anti-Money Laundering and Countering Financing of Terrorism

        On 22 February 2019, FATF issued a public statement recognising the need
        to adequately mitigate the ML and TF risks associated with digital asset
        activities.  As per the statement, FATF proposed more details relating
        to the regulation and supervision/monitoring of virtual assets (“VAs”)
        and virtual asset services providers (“VASPs”)  by way of its (Draft)
        Interpretive Note to Recommendation 15, “New technologies”.
      - "REGULATORY REQUIREMENTS FOR AUTHORISED PERSONS ENGAGED IN REGULATED ACTIVITIES IN RELATION TO VIRTUAL ASSETS\nTechnology Governance and Controls\nWhen complying with GEN Rule 3.3 and COBS Rule 17.5, Authorised Persons should have due regard to the following key areas from a technology perspective:\n\na)\tCareful maintenance and development of systems and architecture (e.g., code version control, implementation of updates, issue resolution, and regular internal and third party testing);\n\nb)\tSecurity measures and procedures for the safe storage and transmission of data;\n\nc)\tBusiness continuity and Client engagement planning in the event of both planned and unplanned system outages;\n\nd)\tProcesses and procedures specifying management of personnel and decision-making by qualified staff; and\n\ne)\tProcedures for the creation and management of services, interfaces and channels provided by or to third parties (as recipients and providers of data or services).\n"
      - >-
        Other stakeholders. The Directors should include in the annual report an
        explanation of the basis on which the Reporting Entity generates or
        preserves value over the longer term (the business model) and the
        strategy for delivering the objectives of the Reporting Entity.
  - source_sentence: >-
      Could you elaborate on the types of 'relevant events' that must be
      reported by Fund Administrators, particularly those which might undermine
      their ability to fulfill their duties as per Rule 17.1.5(d)?
    sentences:
      - "The Regulator would expect any agreement required under this Rule ‎17.1.5 to include as a minimum the following provisions:\n(a)\tunambiguous descriptions and definitions of the activities and functions to be provided by the Fund Administrator and the duties to be performed by both parties;\n(b)\tan agreed standard in respect of resources and services supported as necessary by performance measures in accordance with the applicable legislation;\n(c)\tthe requirement for regular detailed reporting to a specified frequency from the Fund Administrator in respect of its duties and activities;\n(d)\tprovisions relating to the reporting of relevant events such as technological changes or error reporting and, in particular, any event which undermines the ability of the Fund Administrator to fulfil its duties;\n(e)\tthe requirement for an annual review (at a minimum) of the performance of the functions by the Fund Administrator; and\n(f)\tprovisions relating to records and adequate access by the Foreign Fund Manager, the Fund's auditor or any other Persons providing control or risk management functions for the Fund, as required by the Foreign Fund Manager or applicable laws to that Fund."
      - "A Relevant Person which is part of a Group must ensure that it:\n(a)\thas developed and implemented policies and procedures for the sharing of information between Group entities, including the sharing of information relating to CDD and money laundering risks;\n(b)\thas in place adequate safeguards on the confidentiality and use of information exchanged between Group entities, including consideration of relevant data protection legislation;\n(c)\tremains aware of the money laundering risks of the Group as a whole and of its exposure to the Group and takes active steps to mitigate such risks;\n(d)\tcontributes to a Group-wide risk assessment to identify and assess money laundering risks for the Group; and\n(e)\tprovides its Group-wide compliance, audit and AML/TFS functions with customer account and Transaction information from its Branches and Subsidiaries when necessary for AML/TFS purposes."
      - >-

        There are two methods for calculating the Equity Risk Capital
        Requirement: the standard method and the simplified method. The standard
        method requires two separate calculations. The first is Specific Risk
        and the second is General Market Risk. The simplified method is easier
        to calculate but usually results in a higher Capital Requirement than
        the standard method. In addition, Authorised Persons must calculate an
        Interest Rate Risk Capital Requirement for a forward, a Future, an
        Option or a company issued Warrant.
  - source_sentence: >-
      Can a Third Party be compelled to provide access to material under Section
      255 if that material is relevant to an issue that identifies the Third
      Party?
    sentences:
      - >-
        This Chapter deals with the regulatory requirements arising out of the
        need for Authorised Persons to carry out a self assessment of their risk
        which can be reviewed and assessed by the Regulator. This Chapter
        details the Rules stipulating the need to complete internal risk
        assessments by Authorised Persons in defined frequencies and the
        Regulator's role in reviewing the results of such assessments. In the
        case of Authorised Persons facing financial risks, the requirements in
        this Chapter mandate completion of an Internal Capital Adequacy
        Assessment Process. The Regulator will review the results of such
        internal risk assessments. This Chapter also sets out how the Regulator
        may impose an additional Capital Requirement on a firm specific basis in
        addition to the minimum requirement specified in Chapter 3 of these
        Rules to address higher-than-normal risk.
      - >-
        Accounting Records must be maintained by an Authorised Person and
        Recognised Body such as to enable its Governing Body to ensure that any
        financial statements prepared by the Authorised Person or Recognised
        Body comply with the Regulations and Rules.
      - >-
        Section ‎255 applies to a Third Party as it applies to the person to
        whom the notice to which this section applies was given, in so far as
        the material to which access must be given under that section relates to
        the matter which identifies the Third Party.
  - source_sentence: >-
      What is the immediate action required by an Authorised Person or
      Recognised Body upon discovering that an Employee may have committed a
      fraud against a Customer?
    sentences:
      - "Fraud and errors. Each Authorised Person and Recognised Body must notify the Regulator immediately if one of the following events arises in relation to its activities in or from the ADGM:\n(a)\tit becomes aware that an Employee may have committed a fraud against one of its Customers;\n(b)\ta serious fraud has been committed against it;\n(c)\tit has reason to believe that a Person is acting with intent to commit a serious fraud against it;\n(d)\tit identifies significant irregularities in its accounting or other records, whether or not there is evidence of fraud; or\n(e)\tit suspects that one of its Employees who is Connected with the Authorised Person or Recognised Body's Regulated Activities may be guilty of serious misconduct concerning his honesty or integrity.\n"
      - "If a Relevant Person acquires another business, either in whole or in substantial part, the Regulator would permit the Relevant Person to rely on the CDD conducted by the business it is acquiring, but would expect the Relevant Person to have done the following:\n(a)\tas part of its due diligence for the acquisition, to have taken a reasonable sample of the prospective customers to assess the quality of the CDD undertaken; and\n(b)\tto have undertaken CDD on all the customers to cover any deficiencies identified in (a) as soon as possible following the acquisition, prioritising high-risk customers."
      - >
        Additionally, given their heavy dependence on collecting and processing
        client data and the risks of cyberattacks to their automated and largely
        digital mode of operations, Digital Investment Managers must also put in
        place robust data security policies and systems to ensure compliance
        with all relevant data protection regulations, including the ADGM’s Data
        Protection Regulations and, as appropriate, PRU 6.6 – 6.9.
  - source_sentence: >-
      Are there any anticipated changes to the COBS Rule 17.3 / MIR Rule 3.2.1
      that Authorised Persons should be preparing for in the near future? If so,
      what is the expected timeline for these changes to take effect?
    sentences:
      - "A Relevant Person must ensure that its MLRO implements and has oversight of and is responsible for the following matters:\n(a)\tthe day-to-day operations for compliance by the Relevant Person with its AML/TFS policies, procedures, systems and controls;\n(b)\tacting as the point of contact to receive internal notifications of suspicious activity from the Relevant Person's Employees under Rule ‎14.2.2;\n(c)\ttaking appropriate action under Rule ‎14.3.1 following receipt of a notification from an Employee;\n(d)\tmaking, in accordance with Federal AML Legislation, Suspicious Activity/Transaction Reports;\n(e)\tacting as the point of contact within the Relevant Person for competent U.A.E. authorities and the Regulator regarding money laundering issues;\n(f)\tresponding promptly to any request for information made by competent U.A.E. authorities or the Regulator;\n(g)\treceiving and acting upon any relevant findings, recommendations, guidance, directives, resolutions, Sanctions, notices or other conclusions described in Chapter ‎11; and\n(h)\testablishing and maintaining an appropriate money laundering training programme and adequate awareness arrangements under Chapter ‎13."
      - >
        REGULATORY REQUIREMENTS FOR AUTHORISED PERSONS ENGAGED IN REGULATED
        ACTIVITIES IN RELATION TO VIRTUAL ASSETS

        Capital Requirements

        When applying COBS Rule 17.3 / MIR Rule 3.2.1 to an Authorised Person,
        the FSRA will apply proportionality in considering whether any
        additional capital buffer must be held, based on the size, scope,
        complexity and nature of the activities and operations of the Authorised
        Person and, if so, the appropriate amount of regulatory capital required
        as an additional buffer. An Authorised Person that the FSRA considers to
        be high risk may attract higher regulatory capital requirements.
      - "In exceptional circumstances, where the Bail-in Tool is applied, the Regulator may exclude or partially exclude certain liabilities from the application of the Write Down or Conversion Power where—\n(a)\tit is not possible to bail-in that liability within a reasonable time despite the reasonable efforts of the Regulator;\n(b)\tthe exclusion is strictly necessary and is proportionate to achieve the continuity of Critical Functions and Core Business Lines in a manner that maintains the ability of the Institution in Resolution to continue key operations, services and transactions;\n(c)\tthe exclusion is strictly necessary and proportionate to avoid giving rise to widespread contagion, in particular as regards Deposits and Eligible Deposits which would severely disrupt the functioning of financial markets, including financial market infrastructures, in a manner that could cause broader financial instability; or\n(d)\tthe application of the Bail-in Tool to those liabilities would cause a destruction of value such that the losses borne by other creditors would be higher than if those liabilities were excluded from bail-in."
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
  - dot_accuracy@1
  - dot_accuracy@3
  - dot_accuracy@5
  - dot_accuracy@10
  - dot_precision@1
  - dot_precision@3
  - dot_precision@5
  - dot_precision@10
  - dot_recall@1
  - dot_recall@3
  - dot_recall@5
  - dot_recall@10
  - dot_ndcg@10
  - dot_mrr@10
  - dot_map@100
model-index:
  - name: SentenceTransformer based on dunzhang/stella_en_1.5B_v5
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: cosine_accuracy@1
            value: 0.6233859397417504
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.7636298421807748
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8113342898134863
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.8558106169296987
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6233859397417504
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2687709230033477
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.17568149210903872
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09533715925394547
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.5457735533237685
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.6823290291726446
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.7313605930176948
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.7834947393591583
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.6892546786573623
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7027094577668452
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.6454912452493724
            name: Cosine Map@100
          - type: dot_accuracy@1
            value: 0.3446915351506456
            name: Dot Accuracy@1
          - type: dot_accuracy@3
            value: 0.5656384505021521
            name: Dot Accuracy@3
          - type: dot_accuracy@5
            value: 0.6639167862266858
            name: Dot Accuracy@5
          - type: dot_accuracy@10
            value: 0.7786944045911047
            name: Dot Accuracy@10
          - type: dot_precision@1
            value: 0.3446915351506456
            name: Dot Precision@1
          - type: dot_precision@3
            value: 0.19548063127690102
            name: Dot Precision@3
          - type: dot_precision@5
            value: 0.14031563845050213
            name: Dot Precision@5
          - type: dot_precision@10
            value: 0.0854734576757532
            name: Dot Precision@10
          - type: dot_recall@1
            value: 0.3028813964610234
            name: Dot Recall@1
          - type: dot_recall@3
            value: 0.49997010999521757
            name: Dot Recall@3
          - type: dot_recall@5
            value: 0.5915172166427547
            name: Dot Recall@5
          - type: dot_recall@10
            value: 0.7070540411286466
            name: Dot Recall@10
          - type: dot_ndcg@10
            value: 0.5127009608010437
            name: Dot Ndcg@10
          - type: dot_mrr@10
            value: 0.4801890471635807
            name: Dot Mrr@10
          - type: dot_map@100
            value: 0.4463977594142586
            name: Dot Map@100

SentenceTransformer based on dunzhang/stella_en_1.5B_v5

This is a sentence-transformers model finetuned from dunzhang/stella_en_1.5B_v5. It maps sentences & paragraphs to a 1024-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: dunzhang/stella_en_1.5B_v5
Maximum Sequence Length: 512 tokens
Output Dimensionality: 1024 tokens
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': 512, 'do_lower_case': False}) with Transformer model: Qwen2Model 
  (1): Pooling({'word_embedding_dimension': 1536, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Dense({'in_features': 1536, 'out_features': 1024, 'bias': True, 'activation_function': 'torch.nn.modules.linear.Identity'})
)

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("DrishtiSharma/stella_en_1.5B_v5-obliqa-5-epochs")
# Run inference
sentences = [
    'Are there any anticipated changes to the COBS Rule 17.3 / MIR Rule 3.2.1 that Authorised Persons should be preparing for in the near future? If so, what is the expected timeline for these changes to take effect?',
    'REGULATORY REQUIREMENTS FOR AUTHORISED PERSONS ENGAGED IN REGULATED ACTIVITIES IN RELATION TO VIRTUAL ASSETS\nCapital Requirements\nWhen applying COBS Rule 17.3 / MIR Rule 3.2.1 to an Authorised Person, the FSRA will apply proportionality in considering whether any additional capital buffer must be held, based on the size, scope, complexity and nature of the activities and operations of the Authorised Person and, if so, the appropriate amount of regulatory capital required as an additional buffer. An Authorised Person that the FSRA considers to be high risk may attract higher regulatory capital requirements.\n',
    'In exceptional circumstances, where the Bail-in Tool is applied, the Regulator may exclude or partially exclude certain liabilities from the application of the Write Down or Conversion Power where—\n(a)\tit is not possible to bail-in that liability within a reasonable time despite the reasonable efforts of the Regulator;\n(b)\tthe exclusion is strictly necessary and is proportionate to achieve the continuity of Critical Functions and Core Business Lines in a manner that maintains the ability of the Institution in Resolution to continue key operations, services and transactions;\n(c)\tthe exclusion is strictly necessary and proportionate to avoid giving rise to widespread contagion, in particular as regards Deposits and Eligible Deposits which would severely disrupt the functioning of financial markets, including financial market infrastructures, in a manner that could cause broader financial instability; or\n(d)\tthe application of the Bail-in Tool to those liabilities would cause a destruction of value such that the losses borne by other creditors would be higher than if those liabilities were excluded from bail-in.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# 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	0.6234
cosine_accuracy@3	0.7636
cosine_accuracy@5	0.8113
cosine_accuracy@10	0.8558
cosine_precision@1	0.6234
cosine_precision@3	0.2688
cosine_precision@5	0.1757
cosine_precision@10	0.0953
cosine_recall@1	0.5458
cosine_recall@3	0.6823
cosine_recall@5	0.7314
cosine_recall@10	0.7835
cosine_ndcg@10	0.6893
cosine_mrr@10	0.7027
cosine_map@100	0.6455
dot_accuracy@1	0.3447
dot_accuracy@3	0.5656
dot_accuracy@5	0.6639
dot_accuracy@10	0.7787
dot_precision@1	0.3447
dot_precision@3	0.1955
dot_precision@5	0.1403
dot_precision@10	0.0855
dot_recall@1	0.3029
dot_recall@3	0.5
dot_recall@5	0.5915
dot_recall@10	0.7071
dot_ndcg@10	0.5127
dot_mrr@10	0.4802
dot_map@100	0.4464

Training Details

Training Dataset

Unnamed Dataset

Size: 22,291 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: 16 tokens
mean: 33.53 tokens
max: 71 tokens

min: 15 tokens
mean: 118.07 tokens
max: 512 tokens

	sentence_0	sentence_1
type	string	string
details	min: 16 tokens mean: 33.53 tokens max: 71 tokens	min: 15 tokens mean: 118.07 tokens max: 512 tokens

Samples:

sentence_0	sentence_1
`What constitutes a "sufficiently advanced stage of development" for a FinTech Proposal to qualify for a live test under the RegLab framework, as mentioned in criterion (c)?`	Evaluation Criteria. To qualify for authorisation under the RegLab framework, the applicant must demonstrate how it satisfies the following evaluation criteria: (a) the FinTech Proposal promotes FinTech innovation, in terms of the business application and deployment model of the technology. (b) the FinTech Proposal has the potential to: i. promote significant growth, efficiency or competition in the financial sector; ii. promote better risk management solutions and regulatory outcomes for the financial industry; or iii. improve the choices and welfare of clients. (c) the FinTech Proposal is at a sufficiently advanced stage of development to mount a live test. (d) the FinTech Proposal can be deployed in the ADGM and the UAE on a broader scale or contribute to the development of ADGM as a financial centre, and, if so, how the applicant intends to do so on completion of the validity period.
`Are there any upcoming regulatory changes that Authorised Persons should be aware of regarding the handling or classification of Virtual Assets within the ADGM?`	`CONCEPTS RELATING TO THE DISCLOSURE OF PETROLEUM ACTIVITIES Petroleum Projects and materiality If a Petroleum Reporting Entity discloses estimates that it viewed as material at the time of disclosure, but subsequently forms a view that they are no longer material, the FSRA expects the Petroleum Reporting Entity to make a further disclosure providing the clear rationale for the change view on materiality. Such reasoning would generally follow the considerations outlined in paragraph 24 above.`
`What are the ADGM's requirements for VC Managers regarding the periodic assessment and audit of their compliance frameworks, and who is qualified to conduct such assessments?`	Principle 1 – A Robust and Transparent Risk-Based Regulatory Framework. The framework encompasses a suite of regulations, activity-specific rules and supporting guidance that delivers protection to investors, maintains market integrity and future-proofs against financial stability risks. In particular, it introduces a clear taxonomy defining VAs as commodities within the wider Digital Asset universe and requires the licensing of entities engaged in regulated activities that use VAs within ADGM.

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim"
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 10
per_device_eval_batch_size: 10
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: 10
per_device_eval_batch_size: 10
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: 3
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: False
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: False
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
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
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss	cosine_map@100
0.0897	200	-	0.5597
0.1794	400	-	0.5674
0.2242	500	0.7416	-
0.2691	600	-	0.4684
0.3587	800	-	0.5593
0.4484	1000	0.6613	0.5502
0.5381	1200	-	0.5740
0.6278	1400	-	0.5398
0.6726	1500	0.5382	-
0.7175	1600	-	0.5820
0.8072	1800	-	0.5770
0.8969	2000	0.4959	0.5834
0.9865	2200	-	0.5382
1.0	2230	-	0.3223
1.0762	2400	-	0.5532
1.1211	2500	0.3796	-
1.1659	2600	-	0.5817
1.2556	2800	-	0.5929
1.3453	3000	0.367	0.5937
1.4350	3200	-	0.5907
1.5247	3400	-	0.6024
1.5695	3500	0.2877	-
1.6143	3600	-	0.6006
1.7040	3800	-	0.6131
1.7937	4000	0.2818	0.6167
1.8834	4200	-	0.6040
1.9731	4400	-	0.6144
2.0	4460	-	0.6225
2.0179	4500	0.2529	-
2.0628	4600	-	0.6196
2.1525	4800	-	0.6222
2.2422	5000	0.1409	0.6278
2.3318	5200	-	0.6337
2.4215	5400	-	0.6409
2.4664	5500	0.1213	-
2.5112	5600	-	0.6424
2.6009	5800	-	0.6412
2.6906	6000	0.1218	0.6432
2.7803	6200	-	0.6456
2.8700	6400	-	0.6446
2.9148	6500	0.1247	-
2.9596	6600	-	0.6458
3.0	6690	-	0.6455

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.1.1
Transformers: 4.45.2
PyTorch: 2.1.0+cu118
Accelerate: 1.2.0.dev0
Datasets: 3.1.0
Tokenizers: 0.20.3

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",
}

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}
}