metadata

base_model: sentence-transformers/all-mpnet-base-v2
library_name: sentence-transformers
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:152151
  - loss:HardMultipleNegativesRankingLoss
  - loss:CachedMultipleNegativesSymmetricRankingLoss
widget:
  - source_sentence: >-
      Use arc welding techniques to make welds in conditions of very high
      pressure, usually in an underwater dry chamber such as a diving bell.
      Compensate for the negative consequences of high pressure on a weld, such
      as the shorter and less steady welding arc.
    sentences:
      - skill_skill
      - weld in hyperbaric conditions
      - human-robot collaboration
  - source_sentence: >-
      Carry out mineral processing operations, which aim to separate valuable
      minerals from waste rock or grout. Oversee and implement processes such as
      samping, analysis and most importantly the electrostatic separation
      process, which separates valuable materials from mineral ore.
    sentences:
      - internet governance
      - implement mineral processes
      - skill_skill
  - source_sentence: >-
      looking for a pest control technician with strong knowledge in
      preventative measures to minimize pest populations A successful candidate
      will have experience in cryopreservation techniques as well as laboratory
      protocols
    sentences:
      - cryopreservation
      - food preservation
      - skill_sentence
  - source_sentence: >-
      Candidates with experience using popular balance sheet software are
      encouraged to apply for our accounting position. We are looking for a
      cargo handling expert who can maximize efficiency on our shipping vessels.
    sentences:
      - skill_sentence
      - perform balance sheet operations
      - promote inclusion
  - source_sentence: >-
      Must have the ability to read and interpret schematics and effectively
      install and calibrate lift governors to ensure compliance with safety
      standards. The ideal candidate must have an ear for identifying music with
      commercial potential and understand the current market trends.
    sentences:
      - prepare credit reports
      - install lift governor
      - skill_sentence

SentenceTransformer based on sentence-transformers/all-mpnet-base-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-mpnet-base-v2 on the skill_sentence and skill_skill datasets. 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: sentence-transformers/all-mpnet-base-v2
Maximum Sequence Length: 96 tokens
Output Dimensionality: 768 tokens
Similarity Function: Cosine Similarity
Training Datasets:
- skill_sentence
- skill_skill

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': 96, 'do_lower_case': False}) with Transformer model: MPNetModel 
  (1): SmartTokenPooling({'word_embedding_dimension': 768, 'window_size': -1})
)

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("jensjorisdecorte/ConTeXT-Skill-Extraction-base")
# Run inference
sentences = [
    'Must have the ability to read and interpret schematics and effectively install and calibrate lift governors to ensure compliance with safety standards. The ideal candidate must have an ear for identifying music with commercial potential and understand the current market trends.',
    'install lift governor',
    'skill_sentence',
]
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 Datasets

skill_sentence

Dataset: skill_sentence
Size: 138,260 training samples
Columns: anchor, positive, and type

Approximate statistics based on the first 1000 samples:

	anchor	positive	type
type	string	string	string
details	min: 9 tokens mean: 35.67 tokens max: 63 tokens	min: 3 tokens mean: 6.12 tokens max: 15 tokens	min: 5 tokens mean: 5.0 tokens max: 5 tokens

Samples:

anchor	positive	type
`duties for this role will include conducting water chemistry analysis and managing the laboratory. seeking a seasoned print manufacturing manager with knowledge of printing materials, processes and equipment.`	`water chemistry analysis`	`skill_sentence`
`divers must understand how to calculate dive times and limits to ensure they return safely. We are searching for a multimedia software expert with experience in sound, lighting and recording software.`	`comply with the planned time for the depth of the dive`	`skill_sentence`
`A successful candidate will possess the ability to calibrate laboratory equipment according to industry standards. we are seeking a candidate with experience in preparing government funding dossiers`	`prepare government funding dossiers`	`skill_sentence`

Loss: custom_losses.HardMultipleNegativesRankingLoss with these parameters:
```
{
    "scale": 20,
    "similarity_fct": "<lambda>"
}
```

skill_skill

Dataset: skill_skill
Size: 13,891 training samples
Columns: anchor, positive, and type

Approximate statistics based on the first 1000 samples:

	anchor	positive	type
type	string	string	string
details	min: 6 tokens mean: 29.09 tokens max: 96 tokens	min: 3 tokens mean: 6.24 tokens max: 16 tokens	min: 5 tokens mean: 5.0 tokens max: 5 tokens

Samples:

anchor	positive	type
`Adapt and move set pieces during rehearsals and live performances.`	`adapt sets`	`skill_skill`
`Prepare bread and bread products such as sandwiches for consumption.`	`prepare bread products`	`skill_skill`
`The strategies, methods and techniques that increase the organisation's capacity to protect and sustain the services and operations that fulfil the organisational mission and create lasting values by effectively addressing the combined issues of security, preparedness, risk and disaster recovery.`	`organisational resilience`	`skill_skill`

Loss: CachedMultipleNegativesSymmetricRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "mini_batch_size": 64
}

Training Hyperparameters

Non-Default Hyperparameters

overwrite_output_dir: True
eval_strategy: steps
per_device_train_batch_size: 4096
per_device_eval_batch_size: 4096
num_train_epochs: 1
warmup_ratio: 0.1
fp16: True
load_best_model_at_end: True

All Hyperparameters

Click to expand

overwrite_output_dir: True
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 4096
per_device_eval_batch_size: 4096
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: 1.0
num_train_epochs: 1
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: 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: True
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: proportional

Training Logs

Epoch	Step
0.1053	4
0.2105	8
0.3158	12
0.4211	16
0.5263	20
0.6316	24
0.7368	28
0.8421	32
0.9474	36

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.9.19
Sentence Transformers: 3.1.0
Transformers: 4.44.2
PyTorch: 2.4.1+cu118
Accelerate: 0.34.2
Datasets: 3.0.0
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",
}