This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. 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
Base model: sentence-transformers/all-MiniLM-L6-v2
Maximum Sequence Length: 256 tokens
Output Dimensionality: 384 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': 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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("mazej/all-MiniLM-L6-v2_MultipleNegativesRankingLoss_b32_accs8_185348_fine-tuned")
# Run inference
sentences = [
    'Do you need to get a -or- an European ID',
    'When should I use “a” vs “an”? Im writing about concepts in programming languages, and for instance in the Java language, so-called annotations are declared with an @ sign in front of them. When such annotations are referred to in the text, is the @ typically pronounced by the reader, or is it silent? That is, would you write > Use an annotation or > Use a annotation My gut feeling is that the latter variant is better, but then again, Im not a native English speaker (nor a native Java programmer). I would prefer to be explicit, and thus not to write > Use a SuppressWarnings annotation',
    'I would like to collect some census data for every block group in New England. Im primarily interested in median income and median age, and ideally Id like a large .csv file with one row per census block group and one column per variable (e.g. id variables, median_income, etc.). American FactFinder seems almost useless for this kind of thing. Does the census have a bulk data export option?',
]
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]

Evaluation

Metrics

Information Retrieval

Dataset: ir_eval
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@10	0.7608
cosine_precision@10	0.1019
cosine_recall@10	0.7082
cosine_ndcg@10	0.5926
cosine_mrr@10	0.5819
cosine_map@100	0.5486
dot_accuracy@10	0.7608
dot_precision@10	0.1019
dot_recall@10	0.7082
dot_ndcg@10	0.5926
dot_mrr@10	0.5819
dot_map@100	0.5486

Training Details

Training Dataset

Unnamed Dataset

Size: 18,364 training samples
Columns: query and document
Approximate statistics based on the first 1000 samples:
query document
type string string
details
min: 4 tokens
mean: 12.95 tokens
max: 33 tokens
min: 14 tokens
mean: 127.68 tokens
max: 256 tokens

Samples:

query	document
`Extract date from a variable in a different format`	`I am facing following issue: Input: F=date +%Y%m%d echo $F Output: F=20131231 But when trying this(File Name:A.sh): C=date G=$C +%Y%m%d echo $C echo $G Output: A.sh[6]: Tue: not found Tue Dec 31 06:14:12 EST 2013 I want to pass the date in a variable and then extract only date in a specific format.`
`Adding Disqus to wordpress effect SEO`	I am building a website and am trying to decide between writing my own commenting system and using Disqus. One of the main deciding factors is that (obviously), I want comments left on my page to be show on SERPS. However, I remember reading somewhere that Disqus loads comments into a page using AJAX - and therefore the comments are invisible as far as Googlebot and other SE crawlers are concerned. Could someone confirm or refute this? The other question I have is about whether (as a commenter), When I place a comment on another website using Disqus (including any links I may add to my comment), do the links in my comment count as a back link (in other words are they dofollow or nofollow links)?
`cp : short way of copying`	`I often find myself copy-pasting long path in order to create a copy of a file cp /path/to/file/file1 /path/to/file/file1.bkp Is there an alternative utility that will NOT require me to type the path/to/file twice? Something like - nameOfExecutible /path/to/file/file1 bkp Note: I don't want to do a cd to file1s parent directory.`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim" 4}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
gradient_accumulation_steps: 8
num_train_epochs: 1
warmup_ratio: 0.1
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
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: 8
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: 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
use_liger_kernel: False
eval_use_gather_object: False
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	ir_eval_cosine_map@100
0	0	0.5608
0.9895	71	0.5486

Framework Versions

Python: 3.8.18
Sentence Transformers: 3.2.1
Transformers: 4.45.2
PyTorch: 2.4.1
Accelerate: 1.0.1
Datasets: 3.1.0
Tokenizers: 0.20.3

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",
}

MultipleNegativesRankingLoss

bibtex

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

	query	document
type	string	string
details	min: 4 tokens mean: 12.95 tokens max: 33 tokens	min: 14 tokens mean: 127.68 tokens max: 256 tokens