This is a sentence-transformers model finetuned from CALDISS-AAU/DA-BERT_Old_News_V3. It is the first version of a sentence transformer designed to embed texts based on the key actions undertaken in the texts. It is trained to support the project "Run Away" at Aalborg University. It has been trained on sentences from runaway advertisements that have been tagged based on key verbs, grouped into larger clusters of verbs. The key idea is to foreground action.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: CALDISS-AAU/DA-BERT_Old_News_V3
Maximum Sequence Length: 514 tokens
Output Dimensionality: 768 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': 514, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 768, '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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("JohanHeinsen/Run_Away_Action_Embedding_model_v0.1")
# Run inference
sentences = [
    'saa finde vi os beføyet, saavel for at besørge de publiqve Midlers Stkkerhed, som i Anledning af den Nød og Armod hans Undvigelse har foraarsaget hans efterladte Kone og smaa Børn her, at lade ham herved eftelyse og indkalde til at indfinde sig her Byen',
    'thi anmodes enhver som træffer denne Karl at paagribe ham og imod de sædvanlige Indbringerpenge at levere ham til Compagniet.',
    'men ingen veed om ham at sige.',
]
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

Semantic Similarity

Dataset: validation
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.8932
spearman_cosine	0.838

Training Details

Training Dataset

Unnamed Dataset

Size: 83,246 training samples
Columns: sentence_0, sentence_1, and label
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1 label
type string string int
details
min: 6 tokens
mean: 28.53 tokens
max: 176 tokens
min: 4 tokens
mean: 30.7 tokens
max: 514 tokens
0: ~47.20%
1: ~52.80%

Samples:

sentence_0	sentence_1	label
`Han er født i Kjøbenhavn, er 33 Aar gammel, har staaet ved Kongens Regiment som Tambour i 9 Aar, og gaaer derfor skjæbt paa det venstre Bern har tjent Hr. Fabriker Schrøder, og paa Maderup gaard som Gaardskarl, en kort Tid.`	`som tiener paa Hiørnet af gl. Mynt og Svertegaden`	`1`
`efterlyses han herved med Anmodning, at han, af hvem han skulde forekomme, maatte anholdes og mig derom meddeles Underretning`	`Da min Lærredreng Friderich Senggrav er den 12 Aug. undvigt af sin Lære`	`0`
`give det tilkiende hos A. J. Schottlænder, boende i Pilestræde Nr. 11 Litr. Dførste Sal.`	`Niels Pedersen Eistrup, som i 4 Aar har tjent hos mig for Under-Knegt, er fra mig undvigt den 24 December uden at giøre Rigtighed for det ham var betroet`	`0`

Loss: CosineSimilarityLoss with these parameters:
json
1{ 2 "loss_fct": "torch.nn.modules.loss.MSELoss" 3}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 2
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: 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: 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: 2
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}
tp_size: 0
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
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

Training Logs

Epoch	Step	Training Loss	validation_spearman_cosine
0.0961	500	0.1933	-
0.1922	1000	0.1191	-
0.2883	1500	0.1016	-
0.3844	2000	0.0914	-
0.4805	2500	0.0841	-
0.5766	3000	0.079	-
0.6727	3500	0.0757	-
0.7688	4000	0.0732	-
0.8649	4500	0.0686	-
0.9610	5000	0.0665	-
1.0	5203	-	0.8284
1.0571	5500	0.0627	-
1.1532	6000	0.0548	-
1.2493	6500	0.0557	-
1.3454	7000	0.0562	-
1.4415	7500	0.0557	-
1.5376	8000	0.0527	-
1.6337	8500	0.0504	-
1.7298	9000	0.0539	-
1.8259	9500	0.054	-
1.9220	10000	0.0501	-
2.0	10406	-	0.8380

Framework Versions

Python: 3.11.12
Sentence Transformers: 4.1.0
Transformers: 4.51.3
PyTorch: 2.7.0
Accelerate: 1.6.0
Datasets: 2.19.2
Tokenizers: 0.21.1

Citation

BibTeX

Sentence Transformers

bibtex

@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",
}

	sentence_0	sentence_1	label
type	string	string	int
details	min: 6 tokens mean: 28.53 tokens max: 176 tokens	min: 4 tokens mean: 30.7 tokens max: 514 tokens	0: ~47.20% 1: ~52.80%