SentenceTransformer based on dbourget/pb-ds1-48K

This is a sentence-transformers model finetuned from dbourget/pb-ds1-48K. 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: dbourget/pb-ds1-48K
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': False}) with Transformer model: BertModel 
  (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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("dbourget/pb-ds1-48K-philsim")
# Run inference
sentences = [
    'This essay explores the historical and modern perspectives on the Gettier problem, highlighting the connections between this issue, skepticism, and relevance. Through methods such as historical analysis, induction, and deduction, it is found that while contextual theories and varying definitions of knowledge do not fully address skeptical challenges, they can help clarify our understanding of knowledge. Ultimately, embracing subjectivity and intuition can provide insight into what it truly means to claim knowledge.',
    'Objective: In this essay,  I will try to track some historical and modern stages of the discussion on the Gettier problem, and point out the interrelations of the questions that this problem raises for epistemologists, with sceptical arguments, and a so-called problem of relevance. Methods: historical analysis, induction, generalization, deduction, discourse, intuition results: Albeit the contextual theories of knowledge, the use of different definitions of knowledge, and the different ways of the uses of knowledge do not resolve all the issues that the sceptic can put forward, but they can be productive in giving clarity to a concept of knowledge for us. On the other hand, our knowledge will always have an element of intuition and subjectivity, however not equating to epistemic luck and probability.  Significance novelty: the approach to the context in general, not giving up being a Subject may give us a clarity about the sense of what it means to say – “I know”.',
    "Teaching competency in bioethics has been a concern since the field's inception. The first report on the teaching of contemporary bioethics was published in 1976 by The Hastings Center, which concluded that graduate programs were not necessary at the time. However, the report speculated that future developments may require new academic structures for graduate education in bioethics. The creation of a terminal degree in bioethics has its critics, with scholars debating whether bioethics is a discipline with its own methods and theoretical grounding, a multidisciplinary field, or something else entirely. Despite these debates, new bioethics training programs have emerged at all postsecondary levels in the U.S. This essay examines the number and types of programs and degrees in this growing field.",
]
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: sts-dev
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.9378
spearman_cosine	0.8943
pearson_manhattan	0.971
spearman_manhattan	0.8969
pearson_euclidean	0.9711
spearman_euclidean	0.8966
pearson_dot	0.942
spearman_dot	0.8551
pearson_max	0.9711
spearman_max	0.8969

Training Details

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 190
per_device_eval_batch_size: 190
learning_rate: 5e-06
num_train_epochs: 2
warmup_ratio: 0.1
bf16: True
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: 190
per_device_eval_batch_size: 190
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
learning_rate: 5e-06
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 2
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: True
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