README.md

metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - dense
  - generated_from_trainer
  - dataset_size:16688
  - loss:TripletLoss
base_model: BAAI/bge-large-en-v1.5
widget:
  - source_sentence: Where do I select the build type in R³S Modeler?
    sentences:
      - >-
        ## Perform build on(requiresR³S Enterprise) The Perform build on
        drop-down list is not available when running a results workspace. You
        can use the Perform build on drop-down list to specify whether to use
        distributed processing to build (generate, compile, and link) the
        calculations specified in the batch or the model. You can select an
        option from the drop-down list: * Local: build the batch or the model on
        the current machine (the controller) without using distributed
        processing. * Remote: use distributed processing to build the batch or
        the model on one of the remote machines (a worker). This is useful when
        you do not want to occupy the controller with these tasks. If you build
        or run a batch or a model, R³S Modeler uses the value of the Perform
        build on property of the batch or model as the default value of the
        drop-down list. R³S Modeler uses the value of the Perform build on
        drop-down list to set the value of the Perform build on property of the
        batch or the model in the results workspace. You can specify the
        connector to use for distributed processing on the Distribution tab of
        the Options dialog box. For the Remote option to use distributed
        processing, you must use the Microsoft® HPC Pack connector or the Azure®
        Batch connector.
      - >-
        Server hierarchy The Server hierarchy property is a property of the
        following components: - **Development sandbox workspace or approval
        sandbox workspace(requiresR³S Development Manager)**: The Server
        hierarchy property of a development sandbox workspace or an approval
        sandbox workspace shows the name of the sandbox, branch, and library
        with which you associated the workspace. - **Snapshot
        workspace(requiresR³S Development Manager)**: The Server hierarchy
        property of a snapshot workspace shows the name of the changeset, label,
        or sandbox, the branch, and the library from which you created the
        snapshot.
      - >-
        Result grid The result grid of the Analyzer tab of a results workspace
        shows the results at different calculation dates for the current
        variable, the variables that it depends on, and the variables that
        depend on it if these results are available in sample output. The result
        grid is more useful for analyzing layers than data layers, because a
        data layer has only one calculation date, corresponding to the portfolio
        date of the model or model alias. Each scalar numeric and indicator
        variable has a checkbox before its name. Selecting one of these
        checkboxes clears the others. In the graph pane, R³S Modeler graphs the
        results for the variable whose checkbox you select. If events occur in a
        projection step of a layer, R³S Modeler shows these in pink in the
        result grid. To hide the results for the events, select the Hide events
        checkbox. This also stops events from being indicated by the green
        vertical line in the graph pane. If loops occur in a projection step of
        a layer, R³S Modeler shows these in blue in the result grid. To hide the
        results for the loops, select the Hide loops checkbox. Because loop
        variables have no time associated with them, they are never shown in the
        graph pane. Selecting the Hide events checkbox to hide event results or
        the Hide loops checkbox to hide loop results does not affect the
        results; it just hides them in the result grid. With these checkboxes
        selected, it might not be easy to understand the calculation of
        non-portfolio variables that are summed across event dates in the step
        or how the final values of loop variables have been extracted into step
        variables. The dependency diagram still shows all the precedents and is
        not affected by the checkboxes. The result grid shows the variable being
        analyzed in its first row. Precedent variables are shown immediately
        beneath the chosen variable, and dependent variables are shown below
        these. The currently selected date is highlighted with a yellow box.
        Yellow boxes also highlight the variable being analyzed and its
        precedents and dependents. Highlighting a cell in the result grid and
        pressing the Enter key makes the corresponding variable and date the
        subject of the analysis. You can also do this by right-clicking in the
        result grid and choosing Analyze from the context menu. Highlighting a
        cell in the result grid and pressing the Home key makes the
        corresponding variable the subject of the analysis at the layer start
        date. You can also do this by right-clicking in the result grid and choo
  - source_sentence: Are MtF views supported in R³S Modeler?
    sentences:
      - >-
        ## Properties - **MtF views**: - **General**: Name - **Filters**: Filter
        formula - **Data inputs**: File format - **MtF cube**: MtF view type -
        **Auditing**: Last modified - **Sub MtF views**: - **General**: Sub MtF
        view The other properties of a sub MtF view are the same as those of the
        underlying MtF view. - **MtF view variables**: - **General**: Variable -
        **Formula**: Formula - **Auditing**: Last modified Additionally, the
        properties of the variable specified in the Variable property are
        inherited as global properties.
      - >-
        ## Remarks This function acts like the Choose_Life_Table function
        followed by the Reduce_Life_Table function. It avoids the need to use a
        separate life table variable to store the chosen life table before
        reducing it. When you use this function in the formula of a variable,
        that formula can contain nothing outside the call to this function. This
        means that you cannot include the function call as part of a larger
        expression in a single formula. Instead, you can use a variable to call
        the function and then refer to this variable in the formula of another
        variable. The function first uses the character expression to select a
        life table in the workspace by name. R³S Modeler knows how many
        dimensions the life table should have by counting the number of
        arguments. If there are no arguments for the additional dimensions then
        there should be just two dimensions: * Select_Duration should be
        dimension 1 with start position 0 and * Age should be dimension 2 with
        start position 0. If there are no life tables in the workspace with
        these dimensions and, where applicable, the dimension names and start
        positions you specify for dimensions 3, 4 and 5 then a generator error
        occurs. If the life table named in the character expression does not
        have the dimension names and start positions you specify then a runtime
        error occurs. Then the function removes any dimensions in addition to
        the mandatory Select_Duration and Age dimensions by selecting only the
        life table rates corresponding to the specified element position in each
        of the additional dimensions. This is similar to using the Slice
        function repeatedly. This then leaves a 2-dimensional life table for the
        function to reduce further. Suppose that: * The Select_Duration
        dimension has size r (corresponding to a select period of r-1). * The
        Age dimension has size n (corresponding to a maximum age of n-1). *
        Qx(x, t) denotes the select rate for current age x and current select
        duration t (t < r-1). * Qx(x, ) denotes the ultimate rate for current
        age x. * y is an integer that specifies the age at entry. The function
        returns a reduced life table with dimensions: * Select_Duration of size
        1 and start position 0 and * Age of size n and start position 0. If y is
        in the select age range of the life table (greater than or equal to the
        value of the Minimum select age property and less than or equal to the
        value of the Maximum select age property) then the life table rates in
        the reduced life table are: | Select_Duration Age | 0 0 | 0
      - >-
        The main topic 'Layers' has the following related sub-topics: * **Layer
        examples** : The example user workspace includes examples of layers.
  - source_sentence: What data structure is required for the 'Array' argument?
    sentences:
      - >-
        Maximum select age The Maximum select age property is a property of the
        following component: * Life table You can use this mandatory property to
        specify the highest age at entry for which select rates are available in
        the life table. R³S Modeler uses ultimate rates for ages at entry above
        this age. You can specify an integer greater than or equal to 0 and less
        than or equal to 200. The value you specify should be greater than or
        equal to the value of the Minimum select age property. When the size of
        the Select_Duration dimension of the life table is 1, the value you
        specify should be less than or equal to the size of the Age dimension
        minus 1 (that is, the maximum age of the life table), though this
        property makes no difference in this situation, because there are no
        select rates in the life table. When the size of the Select_Duration
        dimension is greater than 1, the value you specify should be less than
        or equal to the size of the Age dimension minus the size of the
        Select_Duration dimension plus 1 (that is, the maximum age of the life
        table plus 1 minus the select period of the life table) to as to give
        enough element positions in the Age dimension for the select rates for
        this maximum select age at entry.
      - >-
        ## Circumstances The formula for the specified variable contains a
        Move_Left or Move_Right function call - say Move_Left(Array_1,
        <Dimension>, Array_2) or Move_Right(Array_1, <Dimension>, Array_2). The
        function call is invalid because the dimension start positions of
        Array_2 (the 'replacement' array to be attached to Array_1) do not all
        match the start positions of the corresponding dimensions of Array_1.
      - >-
        ## Arguments Array | An array variable or expression. Dimension | A
        dimension name of the arrayArray. The dimension must have at least one
        index value that is not blank.
  - source_sentence: >-
      How does R3S Modeler handle non-integer arguments when the data type is
      indicator?
    sentences:
      - >-
        ## Remarks When you use the Range function in a formula or other
        property whose data type is numeric or indicator, all three arguments
        should be numeric or indicator expressions. When the data type of the
        formula or other property is indicator, R³S Modeler rounds values of the
        arguments that are not integers towards zero to give integers. When the
        data type of the formula or other property is date, the first two
        arguments should be date expressions. You can combine sequences of
        values from different calls to the Range and Set functions by separating
        the function calls with a semicolon (;). You can use the Range and Set
        functions in the Formula property of variables in a data view, database
        view or MtF view to produce multiple copies of each data record, which
        can be useful for producing test data.
      - >-
        Pareto (not available in R³S Modeler Lite ) Returns values relating to
        the Pareto distribution.
      - >-
        ## Circumstances Division of an indicator variable by zero has been
        attempted.
  - source_sentence: What inputs does Qx accept?
    sentences:
      - '## Examples q | =Qx(58, , LT) q | =Qx(59+1, 1, LT)'
      - >-
        Invalid Target Variable Message: '<name>' cannot be an array because it
        is the target input range.
      - >-
        ## Circumstances The definition of a user function has no arguments
        defined.
pipeline_tag: sentence-similarity
library_name: sentence-transformers

SentenceTransformer based on BAAI/bge-large-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-large-en-v1.5. 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: BAAI/bge-large-en-v1.5
• Maximum Sequence Length: 384 tokens
• Output Dimensionality: 1024 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': 384, 'do_lower_case': True, 'architecture': 'BertModel'})
  (1): Pooling({'word_embedding_dimension': 1024, '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("dhruvnayee/help_texted_mined_r3s_0810")
# Run inference
sentences = [
    'What inputs does Qx accept?',
    '## Examples q\ue04f\ue052 | =Qx(58, , LT) q\ue028\ue04f\ue053\ue029\ue02a\ue027 | =Qx(59+1, 1, LT)',
    "Invalid Target Variable Message: '<name>' cannot be an array because it is the target input range.",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.8317, 0.8164],
#         [0.8317, 1.0000, 0.4416],
#         [0.8164, 0.4416, 1.0000]])

Training Details

Training Dataset

Unnamed Dataset

• Size: 16,688 training samples
• Columns: sentence_0, sentence_1, and sentence_2
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	sentence_2
  type	string	string	string
  details	min: 7 tokens mean: 13.85 tokens max: 26 tokens	min: 3 tokens mean: 123.19 tokens max: 384 tokens	min: 3 tokens mean: 118.89 tokens max: 384 tokens

• Samples:

  sentence_0	sentence_1	sentence_2
  What are the new features in this release?	## What's new from previous upgrades	What's new in version 1.2 Targeting * This functionality may be used to find the value of an input variable giving the specified value of an output variable, for example for use in profit testing Enhancements to distributed processing * This provides for greater functionality in distributed processing Enhancements to Compare * This extends the functionality of the Compare Page to allow comparison of multiple components Initialization variables * This new component allows greater flexibility and simplicity of coding variables, allowing different definitions at the outset of a projection and during the projection Stochastic processes (not available in R³S Modeler Lite ) * This involves enhancements to the existing stochastic process functionality Results workspaces * This enhances the information provided about related results workspaces on the opening page of a workspace Program Linker and Model Builder * This enables greater ease of adding and moving items within these tools Analyzer e...
  What does this element represent?	Data_Source_Name The Data_Source_Name system variable is a character variable that gives the name of the data source. You can use this system variable in a data source in a data process in the data layer of a model. This system variable is a placeholder variable.	Cannot rerun results workspace Message: It is not possible to rerun a results workspace that contained any model that failed to build.
  How does R3S Modeler create parent program records?	Record_Is_Last_Step The Record_Is_Last_Step system variable is an indicator variable that is 1 if the Record_End_Date system variable for the current program record is a date that is in the current projection step and 0 otherwise. This system variable is not defined for parent program records that R³S Modeler creates solely by aggregating child program records (and does not read from data).	## Circumstances This error occurs when the variable used in a formula does not exist in the workspace (for example, it is not in the Variable Chooser ). For example, if a variable, say, CF_Premium is defined as Prem_Annual * Prob_Surr and the variable Prob_Surr does not exist within the workspace then the above error will occur.

• Loss: TripletLoss with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
      "triplet_margin": 5
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• num_train_epochs: 1
• fp16: True
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• 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: 1
• 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
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.2397	500	4.8389
0.4794	1000	4.7385
0.7191	1500	4.7068
0.9588	2000	4.7199

Framework Versions

• Python: 3.11.11
• Sentence Transformers: 5.1.1
• Transformers: 4.49.0
• PyTorch: 2.5.1+cu124
• Accelerate: 1.3.0
• Datasets: 3.2.0
• Tokenizers: 0.21.0

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",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification},
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}