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

This is a sentence-transformers model finetuned from BAAI/bge-base-en-v1.5. 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: BAAI/bge-base-en-v1.5
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 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': True}) with Transformer model: BertModel 
  (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("datasocietyco/bge-base-en-v1.5-course-recommender-v5")
# Run inference
sentences = [
    'The weather is lovely today.',
    "It's so sunny outside!",
    'He drove to the stadium.',
]
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]

Training Details

Training Dataset

Unnamed Dataset

• Size: 45 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 45 samples:

  	anchor	positive
  type	string	string
  details	min: 143 tokensmean: 178.76 tokensmax: 258 tokens	min: 141 tokensmean: 176.76 tokensmax: 256 tokens

• Samples:

  anchor	positive
  Creating and Querying Data in SQL. This course builds on foundational data skills to teach learners how to effectively manipulate data in Structured Query Language (SQL). By the end of this course, learners will be able to describe database structures, import data into a database, and combine and manipulate data within a single table. Learners will also have gained exposure to working with data of different types, including string, numerical, and temporal.. tags: 'DML', 'analytics', 'ERD', 'SQL', 'functions', 'DQL', 'DDL'. Languages: Course language: TBD. Prerequisites: No prerequisite course required. Target audience: Professionals with limited or no experience in SQL or similar languages. Junior analysts or analysts familiar with other similar programming languages and frameworks.	Course Name:Creating and Querying Data in SQL
  Clustering Categorical and Mixed Data in Python. In this course, learners will prepare data for, implement, and optimize three advanced clustering models in Python while comparing their different use cases. In particular, this course focuses on the suitability of different clustering methods for different kinds of data: numerical, categorical, and mixed. Learners will distinguish between k-modes, mean shift, and k-prototypes models, developing their understanding of when each model will best meet their needs.. tags: 'k-prototypes', 'mean-shift', 'clustering', 'k-modes'. Languages: Course language: Python. Prerequisites: No prerequisite course required. Target audience: This is an introductory level course for data scientists who want to learn to detect and visualize underlying patterns and groups in unlabelled data and how to handle different types of data..	Course Name:Clustering Categorical and Mixed Data in Python
  Hierarchical and Density-Based Clustering in Python. In this course, learners will encounter more sophisticated methods for generating clusters within unlabeled data using Python. The first method, hierarchical clustering, creates easy-to-read, tree branch-like clusters in order of increasing specificity. The second method, DBSCAN (Density-Based Spatial Clustering of Applications with Noise), creates groups based on the concentration of data points within a region, facilitating analysis of irregularly shaped data. By the end of this course, learners will prepare data for, implement, and optimize these models.. tags: 'Hierarchical', 'clustering', 'DBSCAN'. Languages: Course language: Python. Prerequisites: No prerequisite course required. Target audience: This is an introductory level course for data scientists who want to learn to detect and visualize underlying patterns and groups in unlabelled data and how to handle different types of data..	Course Name:Hierarchical and Density-Based Clustering in Python

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 5 evaluation samples
• Columns: anchor and positive
• Approximate statistics based on the first 5 samples:

  	anchor	positive
  type	string	string
  details	min: 167 tokensmean: 211.2 tokensmax: 272 tokens	min: 165 tokensmean: 209.2 tokensmax: 270 tokens

• Samples:

  anchor	positive
  Interactive Visualization with Bokeh in Python. Bokeh is a Python library designed for building complex, multi-layered statistical plots. In this course, learners will discuss the advantages of interactive visualizations using this library. They will first generate a series of simple interactive plots, such as bar charts, scatter plots, and histograms. Then, they will clean and wrangle geospatial data to create an interactive choropleth map. They will also discuss how to export and integrate Bokeh visualizations into web sites as HTML widgets.. tags: 'data story telling', 'analytics', 'interactive visualization', 'visualization', 'python', 'bokeh'. Languages: Course language: Python. Prerequisites: No prerequisite course required. Target audience: Professionals some Python experience who would like to expand their skill set to more advanced Python visualization techniques and tools..	Course Name:Interactive Visualization with Bokeh in Python
  Data Visualization with ggplot2 in R. In this course, learners will take their R skills to the next level by preparing data for exploratory analysis and creating basic, static visualizations. Learners will begin by discussing the purpose of conducting exploratory data analysis (EDA) and best practices for profiling a dataset. Then, using both base R and tidyverse packages, learners will generate bar charts, scatter plots, histograms, and other common visualizations to better understand the shape, structure, and features of a sample dataset.. tags: 'ggplot', 'analytics', 'R', 'visualization', 'tidyverse', 'EDA'. Languages: Course language: TBD. Prerequisites: Prerequisite course required: Data Wrangling in R. Target audience: Creating visualizations is a critical means of exploring data and revealing insights. In this course, learners will take their R skills to the next level by preparing data for exploratory analysis and creating basic, static visualizations. Using both base R and tidyverse packages, learners will generate bar charts, scatter plots, histograms, and other common visualizations to better understand the shape, structure, and features of a sample dataset..	Course Name:Data Visualization with ggplot2 in R
  Outlier Detection for Time Series in Python. While many outlier detection techniques are suitable for general-purpose datasets, time series data involves complex dependencies that, if properly modeled, allow past values to predict future values. In this course, learners will explore how to detect outliers in time series data based on their divergence from predicted values over a particular period. They will begin by developing an Autoregressive Integrated Moving Average (ARIMA) model to extrapolate key statistical properties used in forecasting based on a historical dataset. They will then identify outliers by comparing the model's predictions against actuals as the basis of an anomaly detection model.. tags: 'anomaly detection', 'outlier detection', 'AR', 'ARIMA', 'time series', 'MA'. Languages: Course language: Python. Prerequisites: No prerequisite course required. Target audience: Professionals with some Python experience who would like to expand their skills to learn about various outlier detection techniques.	Course Name:Outlier Detection for Time Series in Python

• 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: 16
• per_device_eval_batch_size: 16
• learning_rate: 3e-06
• max_steps: 24
• warmup_ratio: 0.1
• batch_sampler: no_duplicates

All Hyperparameters

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• 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: 3e-06
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 3.0
• max_steps: 24
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• 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: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Framework Versions

• Python: 3.12.8
• Sentence Transformers: 3.1.1
• Transformers: 4.45.2
• PyTorch: 2.2.2
• Accelerate: 1.2.1
• Datasets: 3.2.0
• 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}
}