finetuned-bge-base-en / README.md

juanpprim

Add new SentenceTransformer model.

823ef65 verified 3 months ago

preview code

raw

history blame contribute delete

24.6 kB

metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:221
  - loss:BatchSemiHardTripletLoss
base_model: BAAI/bge-base-en
widget:
  - source_sentence: |

      Name : Vitality Systems
      Category: Facility Management, Health Services
      Department: Office Administration
      Location: Chicago, IL
      Amount: 347.29
      Card: Office Wellness Initiative
      Trip Name: unknown
    sentences:
      - |

        Name : BizPro Connect
        Category: Data Services & Analytics, Telecommunications
        Department: Executive
        Location: London, UK
        Amount: 1249.67
        Card: Q3 Strategic Enhancement Plan
        Trip Name: unknown
      - |

        Name : Global Insights Group
        Category: Subscriptions & Memberships, Data Services & Analytics
        Department: Marketing
        Location: London, UK
        Amount: 1245.67
        Card: Marketing Intelligence Fund
        Trip Name: unknown
      - |

        Name : Allied Workplace Solutions
        Category: Facility Management, Energy Services
        Department: Office Administration
        Location: New York, NY
        Amount: 861.47
        Card: Monthly Expenses Allocation
        Trip Name: unknown
  - source_sentence: |

      Name : Café Del Mar
      Category: Catering Services, Event Planning
      Department: Sales
      Location: Barcelona, ES
      Amount: 578.29
      Card: Q3 Client Engagement
      Trip Name: unknown
    sentences:
      - |

        Name : TranspoSolutions LLP
        Category: Corporate Travel Analyst, Expense Management Services
        Department: Executive
        Location: New York, NY
        Amount: 629.45
        Card: Strategic Partnership Infrastructure
        Trip Name: Tech Symposium NYC
      - |

        Name : Talent Scout Services
        Category: Professional Services, Recruitment Solutions
        Department: HR
        Location: New York, NY
        Amount: 3200.0
        Card: Recruitment Excellence Fund
        Trip Name: unknown
      - |

        Name : FastLane Transport
        Category: Logistics & Transport, Vehicle Services
        Department: Sales
        Location: Miami, FL
        Amount: 158.25
        Card: Sales Travel Expenses
        Trip Name: unknown
  - source_sentence: |

      Name : CleverCo
      Category: Software & Licenses
      Department: Customer Success
      Location: Amsterdam, Netherlands
      Amount: 2999.99
      Card: Digital Engagement Tools
      Trip Name: unknown
    sentences:
      - |

        Name : Miller & Gartner
        Category: Consulting, Business Expense
        Department: Legal
        Location: Chicago, IL
        Amount: 1500.0
        Card: Legal Fund
        Trip Name: unknown
      - |

        Name : Urban Mobility Solutions
        Category: Transportation Services, Leasing Services
        Department: Executive
        Location: Chicago, IL
        Amount: 1023.45
        Card: Strategic Partnership Building
        Trip Name: Vendor Contract Negotiations
      - |

        Name : Tech Haven Solutions
        Category: Integrated Systems Provider, Custom Hardware Solutions
        Department: IT Operations
        Location: Toronto, Canada
        Amount: 1550.43
        Card: Infrastructure Upgrades Project
        Trip Name: unknown
  - source_sentence: |

      Name : Globex Solutions
      Category: Financial Software, Data Management
      Department: Finance
      Location: New York, NY
      Amount: 1324.57
      Card: Global Revenue Enhancement Initiative
      Trip Name: unknown
    sentences:
      - |

        Name : CloudSync Security
        Category: Cloud Solutions, Cybersecurity Services
        Department: IT Operations
        Location: Dublin, Ireland
        Amount: 1239.45
        Card: Integration Compliance Fund
        Trip Name: unknown
      - |

        Name : Cirrus Insights
        Category: Customer Engagement Platform, SaaS
        Department: Sales
        Location: Austin, TX
        Amount: 1899.99
        Card: Annual Software Licensing Fund
        Trip Name: unknown
      - |

        Name : Innovative Patents Co.
        Category: Intellectual Property Services, Legal Services
        Department: Legal
        Location: New York, NY
        Amount: 3250.0
        Card: Patent Acquisition Fund
        Trip Name: unknown
  - source_sentence: |

      Name : TechSavvy Solutions
      Category: Software Services, Online Subscription
      Department: Engineering
      Location: Austin, TX
      Amount: 1200.0
      Card: Annual Engineering Tools Budget
      Trip Name: unknown
    sentences:
      - |

        Name : Kanzan Solutions
        Category: Consulting Services, Business Advisory
        Department: Legal
        Location: Tokyo, Japan
        Amount: 3900.75
        Card: Quarterly Compliance Review
        Trip Name: unknown
      - |

        Name : Omachi Meitetsu
        Category: Transportation Services, Travel Services
        Department: Sales
        Location: Hakkuba Japan
        Amount: 120.0
        Card: Quarterly Travel Expenses
        Trip Name: unknown
      - |

        Name : Globex Tech Solutions
        Category: Office Equipment Providers, IT Services & Solutions
        Department: IT Operations
        Location: New York, NY
        Amount: 1589.75
        Card: Annual IT Enhancement Budget
        Trip Name: unknown
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy
  - dot_accuracy
  - manhattan_accuracy
  - euclidean_accuracy
  - max_accuracy
model-index:
  - name: SentenceTransformer based on BAAI/bge-base-en
    results:
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: bge base en train
          type: bge-base-en-train
        metrics:
          - type: cosine_accuracy
            value: 0.8280542986425339
            name: Cosine Accuracy
          - type: dot_accuracy
            value: 0.17194570135746606
            name: Dot Accuracy
          - type: manhattan_accuracy
            value: 0.8280542986425339
            name: Manhattan Accuracy
          - type: euclidean_accuracy
            value: 0.8280542986425339
            name: Euclidean Accuracy
          - type: max_accuracy
            value: 0.8280542986425339
            name: Max Accuracy
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: bge base en eval
          type: bge-base-en-eval
        metrics:
          - type: cosine_accuracy
            value: 0.9714285714285714
            name: Cosine Accuracy
          - type: dot_accuracy
            value: 0.02857142857142857
            name: Dot Accuracy
          - type: manhattan_accuracy
            value: 0.9714285714285714
            name: Manhattan Accuracy
          - type: euclidean_accuracy
            value: 0.9714285714285714
            name: Euclidean Accuracy
          - type: max_accuracy
            value: 0.9714285714285714
            name: Max Accuracy

SentenceTransformer based on BAAI/bge-base-en

This is a sentence-transformers model finetuned from BAAI/bge-base-en. 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
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("juanpprim/finetuned-bge-base-en")
# Run inference
sentences = [
    '\nName : TechSavvy Solutions\nCategory: Software Services, Online Subscription\nDepartment: Engineering\nLocation: Austin, TX\nAmount: 1200.0\nCard: Annual Engineering Tools Budget\nTrip Name: unknown\n',
    '\nName : Omachi Meitetsu\nCategory: Transportation Services, Travel Services\nDepartment: Sales\nLocation: Hakkuba Japan\nAmount: 120.0\nCard: Quarterly Travel Expenses\nTrip Name: unknown\n',
    '\nName : Globex Tech Solutions\nCategory: Office Equipment Providers, IT Services & Solutions\nDepartment: IT Operations\nLocation: New York, NY\nAmount: 1589.75\nCard: Annual IT Enhancement Budget\nTrip Name: unknown\n',
]
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

Triplet

Dataset: bge-base-en-train
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.8281
dot_accuracy	0.1719
manhattan_accuracy	0.8281
euclidean_accuracy	0.8281
max_accuracy	0.8281

Triplet

Dataset: bge-base-en-eval
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9714
dot_accuracy	0.0286
manhattan_accuracy	0.9714
euclidean_accuracy	0.9714
max_accuracy	0.9714

Training Details

Training Dataset

Unnamed Dataset

Size: 221 training samples
Columns: sentence and label

Approximate statistics based on the first 221 samples:

	sentence	label
type	string	int
details	min: 32 tokens mean: 39.7 tokens max: 49 tokens	0: ~4.52% 1: ~3.17% 2: ~3.17% 3: ~4.52% 4: ~5.43% 5: ~4.07% 6: ~4.52% 7: ~4.07% 8: ~3.62% 9: ~3.62% 10: ~3.17% 11: ~2.71% 12: ~3.62% 13: ~3.17% 14: ~3.62% 15: ~2.26% 16: ~4.52% 17: ~4.07% 18: ~4.07% 19: ~3.17% 20: ~4.98% 21: ~3.17% 22: ~5.43% 23: ~3.62% 24: ~4.07% 25: ~1.81% 26: ~1.81%

Samples:

sentence	label
`Name : Palace Suites Category: Hotel Accommodation, Event Outsourcing Department: Marketing Location: Amsterdam, NL Amount: 1278.64 Card: Annual Conference Stay Trip Name: 2023 Innovation Summit`	`0`
`Name : BuroPro Services Category: Facilities Management, Maintenance Solutions Department: Office Administration Location: Berlin, Germany Amount: 879.99 Card: Monthly Equipment Oversight Trip Name: unknown`	`1`
`Name : TechXperts Global Category: IT Services, Consulting Department: IT Operations Location: Berlin, Germany Amount: 987.49 Card: Quarterly System Assessment Trip Name: unknown`	`2`

Loss: BatchSemiHardTripletLoss

Evaluation Dataset

Unnamed Dataset

Size: 55 evaluation samples
Columns: sentence and label

Approximate statistics based on the first 55 samples:

	sentence	label
type	string	int
details	min: 34 tokens mean: 39.07 tokens max: 46 tokens	0: ~1.82% 1: ~7.27% 2: ~1.82% 3: ~1.82% 4: ~1.82% 5: ~1.82% 6: ~10.91% 7: ~5.45% 8: ~3.64% 9: ~3.64% 10: ~14.55% 12: ~1.82% 13: ~3.64% 14: ~3.64% 15: ~3.64% 16: ~3.64% 19: ~9.09% 20: ~3.64% 22: ~1.82% 23: ~3.64% 24: ~3.64% 25: ~5.45% 26: ~1.82%

Samples:

sentence	label
`Name : NetSolve Consulting Category: IT Consulting, Infrastructure Solutions Department: IT Operations Location: Berlin, Germany Amount: 892.45 Card: Tech Infrastructure Enhancement Trip Name: unknown`	`2`
`Name : Urban Mobility Solutions Category: Transportation Services, Leasing Services Department: Executive Location: Chicago, IL Amount: 1023.45 Card: Strategic Partnership Building Trip Name: Vendor Contract Negotiations`	`10`
`Name : CloudSync Security Category: Cloud Solutions, Cybersecurity Services Department: IT Operations Location: Dublin, Ireland Amount: 1239.45 Card: Integration Compliance Fund Trip Name: unknown`	`15`

Loss: BatchSemiHardTripletLoss

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
learning_rate: 2e-05
num_train_epochs: 5
warmup_ratio: 0.1
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

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
learning_rate: 2e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 5
max_steps: -1
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
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	bge-base-en-eval_max_accuracy	bge-base-en-train_max_accuracy
0	0	-	0.8281
5.0	70	0.9714	-

Framework Versions

Python: 3.9.22
Sentence Transformers: 3.1.1
Transformers: 4.42.2
PyTorch: 2.7.0+cu126
Accelerate: 1.3.0
Datasets: 3.6.0
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",
}

BatchSemiHardTripletLoss

@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}
}