SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. It maps sentences & paragraphs to a 384-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: sentence-transformers/all-MiniLM-L6-v2
Maximum Sequence Length: 128 tokens
Output Dimensionality: 384 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': 128, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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})
)

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("Daxtra/sbert-summaries-minilm-128-batch")
# Run inference
sentences = [
    "- Senior Business Development Executive with over 3+ years of experience in business development or a related field, focusing on expanding client businesses and maintaining strong relationships with MSPs and Resellers.\n- Responsibilities include client-facing roles, strategic planning, lead generation, CRM management using Pipedrive, market analysis, team collaboration, and reporting.\n- Requires proven experience in client-facing roles and proficiency in lead generation, closing deals, and using CRM tools.\n- Strategic mindset and excellent communication, negotiation, and presentation skills are essential.\n- Bachelor's degree in Business, Marketing, Sales, or a related field.",
    '- Business Development Expert with over 4 years in financial services consulting and B2B cold-calling, focusing on market analysis and stakeholder relationships.\n- Current Role: Business Development Representative at Warehouse Club, achieving a 40% increase in sales and maintaining optimal stock levels.\n- Former Role: Brand Ambassador, leading 300% sales growth through persuasive communication and product demonstrations.\n- Skills: Leadership, communication, analytical, and stakeholder management.\n- Certifications: ILSSI Lean Six Sigma Green Belt, Business Analyst-Transfer Pricing, Taxation, and Financial Advisement.\n- Education: Post-graduate with a Master of Science in Management and Bachelor of Commerce.\n- Interests: Project execution, FMCG, and financial services.',
    '- Senior Data/Informatica Engineer with over 15 years in IT, specializing in Big Data/Cloud and Data Warehousing.\n- Led IDMC migration projects, improving scalability and reducing operational costs.\n- Developed mappings for data integration, using Informatica for ETL tasks and Python for advanced transformations.\n- Skilled in Informatica parameter files, cloud transformations, and IDMC migration.\n- Expertise in Python, Informatica, Teradata, GCP, BigQuery, and other data integration tools.\n- Experience in data migration, performance tuning, and cloud integration.\n- Proficient in UNIX/Linux scripting for data processing.\n- Expertise in handling GCP buckets, BQ, HDFS, and HBase.\n- Strong skills in GIT, JIRA, Control M, BitBucket, Bamboo, and Informatica CICD.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

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

Evaluation

Metrics

Information Retrieval

Dataset: vac-res-matcher
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@10	0.3731
cosine_precision@10	0.0762
cosine_recall@10	0.1153
cosine_ndcg@10	0.1206
cosine_mrr@10	0.1944
cosine_map@10	0.0725

Training Details

Training Dataset

Unnamed Dataset

Size: 149,248 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: 47 tokens
mean: 116.19 tokens
max: 128 tokens

min: 54 tokens
mean: 119.22 tokens
max: 128 tokens

	sentence_0	sentence_1
type	string	string
details	min: 47 tokens mean: 116.19 tokens max: 128 tokens	min: 54 tokens mean: 119.22 tokens max: 128 tokens

Samples:

sentence_0	sentence_1
- Public Relations Account Executive role for graduates with interest in media relations and corporate PR. - Key responsibilities include researching media data, managing media relationships, drafting reports, and coordinating media features. - Support client teams, manage social media, and contribute to SEO. - Require a 2.1 degree from a leading university, preferably in Economics, Finance, Business, English, or Communications/Media. - Strong understanding of financial and professional services industries. - Essential skills: excellent writing, trend analysis, integrity, proactive teamwork, and leadership in account support. - Previous PR experience is desirable.	- Experienced Social Media and Productions Manager with a Master’s in Digital Communication and Marketing. - Led the creation of a multi-channel social media platform, generating Rs 900K annual revenue for a beauty brand. - Managed digital asset creation for Palmolive Color Naturals, overseeing talent acquisition to post-production. - Expertise in market segmentation, social media management, and post-production processes. - Proficient in Microsoft Office, PowerPoint, Excel, and Adobe Suite; holds GCSE and O Levels in Math and Economics. - Bilingual in Urdu and English; advanced knowledge in Advertising, Brand Marketing Strategy, and Public Relations.
- Fire/Safety Senior Sales Executive, requiring either 2-5 years of experience for Sr. Sales Executive or 5+ years for Account Executive. - Responsible for developing sales strategies, managing contractor and end-user relationships, and executing sophisticated deals within established guidelines for fire and life safety in Iowa and Nebraska. - Build scope development, develop proposals, interact with customers, and provide value through communication on product and installation risks. - Coordinate estimating efforts and manage multiple projects. - Develop a market understanding, identify new business opportunities, and position Siemens as a leader. - Spend minimum 50% of time in customer-facing activities and travel 20% for training and business development. - Required: High School Diploma or GED, 2+ years/5+ years experience, working knowledge of fire and life safety systems, and experience with building codes. - Must be 21 years old and hold a valid driver's license. - Preferred experience includes selling to contractors, design services, and experience in vertical markets.	- Experienced banking professional with over 7 years in Payments Coordination and Project Management. - Currently seeking a role in Project Management, with strong skills in accurate and timely payment processing, customer service, and system monitoring. - Proficient in handling customer inquiries, credit card fraud detection, and documentation processes. - Holds a Master's Degree in Business Administration and Management, and a Bachelor's in Business. - Certified Associate in Project Management (CAPM) and licensed insurance agent. - Fluent in English, French, and proficient in ArcGIS, Microsoft Access, and various banking applications.
- Quality Assurance (QA) role, Entry-level, focusing on Java and MySQL skills, with a requirement for a Bachelor's Degree in Information Systems, Computer Science, or related fields. - Essential skills include a methodical and analytical mindset, hands-on software development experience, and basic understanding of software development lifecycle and automation. - Must be able to handle evolving requirements and collaborate effectively with teams to achieve common goals. - Experience in financial industry and working with Big Data, SQL, and programming languages like Java, Groovy, Perl, Python, JavaScript is desired. - Requires proficiency in core Java (0-5 scale), with 0 years of experience necessary; proficiency in Business Analysis is also required (0-5 scale).	- Test Automation Engineer with over 6 years of experience in Agile/Scrum environments, skilled in Java and Selenium WebDriver for web-based application testing. - Developed test automation frameworks using Maven, JUnit, and Page Object Model design pattern. - Proficient in Cucumber BDD features, steps, and runner packages for feature testing, as well as data-driven and cross-browser testing. - Executed RESTful API testing with Postman and REST Assured, and database testing using SQL queries and JDBC. - Expertise in test planning, test scripting, defect tracking, and test reporting using Jira Xray and other tools. - Knowledgeable in Continuous Integration (CI/CD) and mentoring junior QA staff. - Proficient in Java, Selenium, Cucumber, Maven, JUnit, Postman, REST Assured, MS Excel, SQL, and JDBC.

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: 128
per_device_eval_batch_size: 128
num_train_epochs: 1
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: 128
per_device_eval_batch_size: 128
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: 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
eval_use_gather_object: False
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss	vac-res-matcher_cosine_map@10
0.0995	116	-	0.0684
0.1990	232	-	0.0699
0.2985	348	-	0.0711
0.3979	464	-	0.0720
0.4288	500	2.6358	-
0.4974	580	-	0.0697
0.5969	696	-	0.0721
0.6964	812	-	0.0714
0.7959	928	-	0.0717
0.8576	1000	2.373	-
0.8954	1044	-	0.0722
0.9949	1160	-	0.0724
1.0	1166	-	0.0725

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.2.1
Transformers: 4.44.2
PyTorch: 2.4.1+cu121
Accelerate: 0.34.2
Datasets: 3.0.1
Tokenizers: 0.19.1

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",
}

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}
}

Daxtra
/

sbert-summaries-minilm-128-batch