SentenceTransformer

This is a sentence-transformers model trained. It maps sentences & paragraphs to a 384-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: 256 tokens
Output Dimensionality: 384 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': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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})
  (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("minsuas/Misconceptions_1")
# Run inference
sentences = [
    'Subject: Construct Triangle\nConstruct: Construct a triangle using Side-Side-Side\nQuestion: Tom and Katie are arguing about constructing triangles.\n\nTom says you can construct a triangle with lengths  $11 \\mathrm{~cm}, 10 \\mathrm{~cm}$  and  $2 \\mathrm{~cm}$ .\n\nKatie says you can construct a triangle with lengths  $8 \\mathrm{~cm}, 5 \\mathrm{~cm}$  and  $3 \\mathrm{~cm}$ .\n\nWho is correct?\nIncorrect Answer: Neither is correct',
    'Does not realise that the sum of the two shorter sides must be greater than the third side for it to be a possible triangle\nThe passage is discussing a common misconception about the properties required to form a triangle. The misconception is that one might think any three given side lengths can form a triangle. However, for three lengths to actually form a triangle, they must satisfy the triangle inequality theorem. This theorem states that the sum of the lengths of any two sides of a triangle must be greater than the length of the remaining side. This rule must hold true for all three combinations of added side lengths. \n\nTo apply this to the misconception: one does not realize that the sum of the lengths of the two shorter sides must be greater than the length of the longest side to form a possible triangle. This ensures that the sides can actually meet to form a closed figure with three angles.',
    'Draws both angles at the same end of the line when constructing a triangle\nThe misconception described refers to a common error in geometry when students are constructing a triangle based on given angles and a line segment. The mistake is to draw both given angles at the same end of the given line segment. This is incorrect because in a triangle, each angle is located at a different vertex, and each vertex connects two sides. To correctly construct the triangle, each given angle should be drawn at different ends of the line segment (if constructing based on one line segment and two angles) or at vertices defined by the construction steps (if additional sides are given). This ensures that the three angles are positioned to form the corners of the triangle, with each angle at a distinct vertex, thereby creating a proper triangle.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Training Details

Training Dataset

Unnamed Dataset

Size: 17,405 training samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative
type	string	string	string
details	min: 32 tokens mean: 87.49 tokens max: 256 tokens	min: 79 tokens mean: 179.08 tokens max: 256 tokens	min: 75 tokens mean: 180.1 tokens max: 256 tokens

Samples:

anchor	positive	negative
`Subject: Cubics and Reciprocals Construct: Given a positive x value, find the corresponding y value for reciprocal graphs Question: This is a part of the table of values for the equation ( y=\frac{3}{x} ) \begin{tabular}{`	l	l
`Subject: Angle Facts with Parallel Lines Construct: Identify a transversal Question: What is the name given to the red line that intersects the two dashed lines? Incorrect Answer: Parallel`	Does not know the meaning of the word parallel The passage is indicating a misconception related to a math problem, specifically one that involves the concept of "parallel." In mathematics, particularly in geometry, "parallel" refers to lines or planes that are equidistant from each other at every point and never intersect, no matter how far they are extended. A misunderstanding or lack of knowledge about this definition can lead to errors when solving problems that involve parallel lines or planes, such as determining angles or distances. Thus, to correctly interpret and solve problems involving parallel lines or planes, one must understand that they maintain a constant distance from each other and never meet.	Does not know the term transversal The passage is indicating a common pitfall in geometry where a student may not be familiar with the term "transversal." A transversal is a line that passes through two or more other lines, often creating several angles with them. When discussing parallel lines and the angles formed when a transversal intersects them, understanding the term and its implications is crucial. The misconception here is likely that without knowing what a transversal is, a student might struggle to identify the relationships between the angles formed (such as alternate interior angles, corresponding angles, etc.), which are fundamental concepts in solving geometry problems involving parallel lines.
`Subject: Sharing in a Ratio Construct: Divide a quantity into two parts for a given a ratio, where each part is an integer Question: Share ( £360 ) in the ratio ( 2: 7 ) Incorrect Answer: ( £ 180: £ 51 )`	Divides total amount by each side of the ratio instead of dividing by the sum of the parts The misconception described refers to a mistake made when dividing a total amount according to a given ratio. For instance, if someone has to divide $100 in the ratio 2:3, a correct approach would be to first add the parts of the ratio (2+3=5) to find the total number of parts. Then, divide the total amount by this sum ($100 ÷ 5 = $20) to determine the value of one part. This $20 can then be multiplied by each number in the ratio (2 and 3) to correctly distribute the $100. The misconception occurs when someone divides the total amount ($100) by each individual number in the ratio (2 and 3) rather than by the sum of the parts (5). This method would incorrectly distribute the $100, as it does not account for the proportional relationship that the ratio is meant to establish.	Estimates shares of a ratio instead of calculating The passage is discussing a common mistake made in mathematics, particularly when dealing with ratio problems. The misconception lies in estimating the shares or parts of a ratio rather than calculating them accurately. For example, if a problem involves dividing a quantity in the ratio of 2:3, the misconception would be to guess or estimate what parts of the quantity correspond to 2 and 3, instead of using the correct method to find the exact shares. The correct approach involves first adding the parts of the ratio (in this case, 2 + 3 = 5) and then using this sum to calculate each part's exact share of the total quantity. Thus, it's important to calculate each part of the ratio precisely rather than estimating.

Loss: CachedMultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 512
num_train_epochs: 1
lr_scheduler_type: cosine
warmup_ratio: 0.1
fp16: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 512
per_device_eval_batch_size: 8
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: cosine
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: 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
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
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
use_liger_kernel: False
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.3.1
Transformers: 4.47.1
PyTorch: 2.5.1+cu121
Accelerate: 1.2.1
Datasets: 3.2.0
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",
}

CachedMultipleNegativesRankingLoss

@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}