Culture-and-Morality-Lab/psyembedding-bert-large-uncased

SentenceTransformer

This is a sentence-transformers model trained. It maps sentences & paragraphs to a 1024-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
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 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': 512, 'do_lower_case': False, 'architecture': 'BertModel'})
  (1): Pooling({'word_embedding_dimension': 1024, '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("sentence_transformers_model_id")
# Run inference
sentences = [
    'Every American poet feels that the whole responsibility for contemporary poetry has fallen upon his shoulders, that he is a literary aristocracy of one.',
    "I'm not sure why Spike Lee made this train wreck of a movie and conned poor Stevie Wonder into eternally pairing his beautiful music with this theatrical mess. I also resent the way he uses profanity as a part of the normal prose of professional Blacks. The abuse of his hold on ethnic movie goers is a shame. Scenes which seem to be contrived out the blue and have nothing to do with the theme or sub themes, play as if some college kid wrote this. I especially detest the ludicrous scene where the two leads are playfully sparring for no reason at all and the cops come and rough up Snipes. The overacting of the leads makes one feel as if Spike has no respect for his viewers or he has no clue what a movie is all about. The final scene appears to be thrown in to justify the use of a sledge hammer to tack a point in. This movie also supports the myth that all people of culture use the F-word in casual conversation. I am hoping he will realize that the rest of his movies are in the same pool as this one where he is not growing as a film maker. I think his union with Scorcesee in Clockers was a wise move. He should stick to making documentaries like the Four Little Colored Girls. Shock movies do not an Oscar make.",
    'Many thousands of youth have been deprived of the benefit of education thereby, their morals ruined, and talents irretrievably lost to society, for want of cultivation: while two parties have been idly contending who should bestow it.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.6358, 0.5940],
#         [0.6358, 1.0000, 0.4347],
#         [0.5940, 0.4347, 1.0000]])

Evaluation

Metrics

Semantic Similarity

• Dataset: similarity
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.3694
spearman_cosine	0.3901

Training Details

Training Dataset

Unnamed Dataset

• Size: 11,180 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	float
  details	min: 6 tokens mean: 107.53 tokens max: 512 tokens	min: 5 tokens mean: 111.53 tokens max: 512 tokens	min: 0.0 mean: 0.53 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  I AM ANGRY AT YOU BILLJ! YOU GOT PEOPLE BLOCKED FOR AS LONG AS YOU LIVE! I ASKED YOU TO STOP DELETING MY EDITS OR I WILL BLOCK YOU FOR ALl EONS YOU ASSHOLE! WIKIPEDIA IS NOT CENSORED SO STOP REMOVING MY FUCKING MESSAGES OR I WILL BEAT YOU UP SILLY!	The thing is i don't see any shyness from people supporting far right anymore. The life of avg Joe became signicantly shittier after covid and global conflicts. They are very vocal about their distaste. And they blame lefties and immigrants for their problems. So they are vocal and very organized. Also most of the public already act demented and noone remembers all the moronic stuff Trump pulled during his presidency. I don't see much reason for them to be shy about.	0.4082482904638631
  I understand that you may be confused, but you still shouldn't judge someone's sexual identity. Just because they haven't acted on all of their sexual inclinations, doesn't mean that they don't still have those feelings. Accept others as they present themselves.	The head of the Mormon church has married same sex couples in the temple because they were close family. It's all about $$$.	0.5773502691896258
  Ugh there is so many bad decisions by conservative judges that need to be undone.	you say waste a draft pick on Manziel when we have Mallet. that's why I'm telling you to delete your account. You're retarded	0.5773502691896258

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• fp16: True
• 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: 32
• per_device_eval_batch_size: 32
• 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: 3
• 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: 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: 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: None
• hub_always_push: False
• hub_revision: None
• 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
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss	similarity_spearman_cosine
0.0286	10	-	-0.0664
0.0571	20	-	-0.0621
0.0857	30	-	-0.0581
0.1143	40	-	-0.0516
0.1429	50	-	-0.0444
0.1714	60	-	-0.0334
0.2	70	-	-0.0194
0.2286	80	-	-0.0061
0.2571	90	-	0.0177
0.2857	100	-	0.0317
0.3143	110	-	0.0510
0.3429	120	-	0.0667
0.3714	130	-	0.0892
0.4	140	-	0.1206
0.4286	150	-	0.1584
0.4571	160	-	0.1821
0.4857	170	-	0.1716
0.5143	180	-	0.1749
0.5429	190	-	0.2192
0.5714	200	-	0.2473
0.6	210	-	0.2399
0.6286	220	-	0.2419
0.6571	230	-	0.2637
0.6857	240	-	0.2672
0.7143	250	-	0.2754
0.7429	260	-	0.2942
0.7714	270	-	0.3079
0.8	280	-	0.3079
0.8286	290	-	0.3077
0.8571	300	-	0.3012
0.8857	310	-	0.3148
0.9143	320	-	0.3199
0.9429	330	-	0.3306
0.9714	340	-	0.3363
1.0	350	-	0.3419
1.0286	360	-	0.3402
1.0571	370	-	0.3366
1.0857	380	-	0.3402
1.1143	390	-	0.3360
1.1429	400	-	0.3371
1.1714	410	-	0.3536
1.2	420	-	0.3268
1.2286	430	-	0.3443
1.2571	440	-	0.3011
1.2857	450	-	0.3549
1.3143	460	-	0.3321
1.3429	470	-	0.3505
1.3714	480	-	0.3412
1.4	490	-	0.3337
1.4286	500	0.1211	0.3488
1.4571	510	-	0.3486
1.4857	520	-	0.3508
1.5143	530	-	0.3561
1.5429	540	-	0.3592
1.5714	550	-	0.2950
1.6	560	-	0.3287
1.6286	570	-	0.3369
1.6571	580	-	0.3407
1.6857	590	-	0.3283
1.7143	600	-	0.3547
1.7429	610	-	0.3665
1.7714	620	-	0.3459
1.8	630	-	0.3614
1.8286	640	-	0.3514
1.8571	650	-	0.3714
1.8857	660	-	0.3647
1.9143	670	-	0.3601
1.9429	680	-	0.3292
1.9714	690	-	0.3321
2.0	700	-	0.3624
2.0286	710	-	0.3605
2.0571	720	-	0.3702
2.0857	730	-	0.3783
2.1143	740	-	0.3788
2.1429	750	-	0.3813
2.1714	760	-	0.3736
2.2	770	-	0.3762
2.2286	780	-	0.3804
2.2571	790	-	0.3805
2.2857	800	-	0.3755
2.3143	810	-	0.3647
2.3429	820	-	0.3654
2.3714	830	-	0.3767
2.4	840	-	0.3727
2.4286	850	-	0.3824
2.4571	860	-	0.3660
2.4857	870	-	0.3791
2.5143	880	-	0.3723
2.5429	890	-	0.3818
2.5714	900	-	0.3861
2.6	910	-	0.3861
2.6286	920	-	0.3857
2.6571	930	-	0.3825
2.6857	940	-	0.3680
2.7143	950	-	0.3750
2.7429	960	-	0.3815
2.7714	970	-	0.3851
2.8	980	-	0.3879
2.8286	990	-	0.3863
2.8571	1000	0.1033	0.3818
2.8857	1010	-	0.3882
2.9143	1020	-	0.3896
2.9429	1030	-	0.3899
2.9714	1040	-	0.3901

Framework Versions

• Python: 3.11.9
• Sentence Transformers: 5.1.0
• Transformers: 4.53.3
• PyTorch: 2.5.1
• Accelerate: 1.10.0
• Datasets: 2.14.4
• Tokenizers: 0.21.0

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",
}