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: 512 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': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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("Adi-0-0-Gupta/Embedding-v2")
# Run inference
sentences = [
    'Tag: vegan pad thai\n\nCook time of vegan pad thai based on different serving sizes: Serving 1 - 20 mins, Serving 2 - 25 mins, Serving 3 - 30 mins, Serving 4 - 35 mins',
    'Tag: vegan pad thai\n\nWhat’s the expected cook time for vegan pad thai?',
    'Tag: scrambled eggs with veggies\n\nWhat dietary classifications suit scrambled eggs with veggies?',
]
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: dim_384
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9688
cosine_accuracy@3	0.9701
cosine_accuracy@5	0.9748
cosine_accuracy@10	0.9947
cosine_precision@1	0.9688
cosine_precision@3	0.847
cosine_precision@5	0.8015
cosine_precision@10	0.4412
cosine_recall@1	0.3286
cosine_recall@3	0.6209
cosine_recall@5	0.8939
cosine_recall@10	0.9605
cosine_ndcg@10	0.9593
cosine_mrr@10	0.9734
cosine_map@100	0.954

Information Retrieval

Dataset: dim_256
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.968
cosine_accuracy@3	0.9693
cosine_accuracy@5	0.9752
cosine_accuracy@10	0.9949
cosine_precision@1	0.968
cosine_precision@3	0.8459
cosine_precision@5	0.7993
cosine_precision@10	0.4392
cosine_recall@1	0.3284
cosine_recall@3	0.6204
cosine_recall@5	0.8919
cosine_recall@10	0.9569
cosine_ndcg@10	0.9567
cosine_mrr@10	0.9727
cosine_map@100	0.9512

Information Retrieval

Dataset: dim_128
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9695
cosine_accuracy@3	0.9705
cosine_accuracy@5	0.9761
cosine_accuracy@10	0.9949
cosine_precision@1	0.9695
cosine_precision@3	0.8471
cosine_precision@5	0.7985
cosine_precision@10	0.438
cosine_recall@1	0.3287
cosine_recall@3	0.6211
cosine_recall@5	0.8913
cosine_recall@10	0.955
cosine_ndcg@10	0.9559
cosine_mrr@10	0.9739
cosine_map@100	0.95

Information Retrieval

Dataset: dim_64
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9699
cosine_accuracy@3	0.972
cosine_accuracy@5	0.9772
cosine_accuracy@10	0.9938
cosine_precision@1	0.9699
cosine_precision@3	0.8473
cosine_precision@5	0.7966
cosine_precision@10	0.434
cosine_recall@1	0.3288
cosine_recall@3	0.6213
cosine_recall@5	0.8895
cosine_recall@10	0.9478
cosine_ndcg@10	0.9517
cosine_mrr@10	0.9744
cosine_map@100	0.946

Information Retrieval

Dataset: dim_32
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.969
cosine_accuracy@3	0.9708
cosine_accuracy@5	0.9769
cosine_accuracy@10	0.993
cosine_precision@1	0.969
cosine_precision@3	0.8465
cosine_precision@5	0.794
cosine_precision@10	0.4319
cosine_recall@1	0.3286
cosine_recall@3	0.6208
cosine_recall@5	0.8871
cosine_recall@10	0.9441
cosine_ndcg@10	0.949
cosine_mrr@10	0.9735
cosine_map@100	0.9417

Training Details

Training Dataset

Unnamed Dataset

Size: 42,333 training samples
Columns: positive and anchor
Approximate statistics based on the first 1000 samples:
positive anchor
type string string
details
min: 17 tokens
mean: 71.13 tokens
max: 433 tokens

min: 14 tokens
mean: 22.97 tokens
max: 41 tokens

	positive	anchor
type	string	string
details	min: 17 tokens mean: 71.13 tokens max: 433 tokens	min: 14 tokens mean: 22.97 tokens max: 41 tokens

Samples:

positive	anchor
`Tag: beef and broccoli A small description of beef and broccoli: Stir fried broccoli and tender beef strips stir-fried in a rich savory sauce.`	`Tag: beef and broccoli How do you describe beef and broccoli?`
`Tag: beef and broccoli Garnishing tips for beef and broccoli: Best served on it's own or on top of hot rice with chopped scallions!`	`Tag: beef and broccoli What are some classic garnishes for beef and broccoli?`
`Tag: beef and broccoli For beef and broccoli, these dietary tags go well with it: dinner, contains soy, meat recipes, asian american cuisine, lunch, american cuisine, beef recipes, asian cuisine, chinese cuisine, hearty recipes, rice recipes, protein rich recipes, non vegetarian, saucy recipes, stir fry recipes, healthy recipes`	`Tag: beef and broccoli What dietary labels suit beef and broccoli?`

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "MultipleNegativesRankingLoss",
    "matryoshka_dims": [
        384,
        256,
        128,
        64,
        32
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
gradient_accumulation_steps: 16
learning_rate: 2e-05
num_train_epochs: 100
lr_scheduler_type: constant
warmup_ratio: 0.1
bf16: True
tf32: True
load_best_model_at_end: True
optim: adamw_torch_fused
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: 32
per_device_eval_batch_size: 32
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 16
eval_accumulation_steps: None
learning_rate: 2e-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: 100
max_steps: -1
lr_scheduler_type: constant
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: True
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: True
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_fused
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
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	dim_128_cosine_map@100	dim_256_cosine_map@100	dim_32_cosine_map@100	dim_384_cosine_map@100	dim_64_cosine_map@100
0.3023	25	2.7893	0.9106	0.9169	0.8833	0.9193	0.9013
0.6047	50	1.6554	0.9061	0.9153	0.8858	0.9199	0.8970
0.9070	75	0.7514	0.9361	0.9382	0.9216	0.9423	0.9292
1.2079	100	1.2044	0.9334	0.9370	0.9186	0.9413	0.9263
1.5102	125	1.4103	0.9312	0.9342	0.9146	0.9382	0.9222
1.8125	150	0.6925	0.9444	0.9463	0.9326	0.9502	0.9385
2.1134	175	0.7937	0.9333	0.9376	0.9196	0.9410	0.9256
2.4157	200	1.3185	0.9321	0.9355	0.9191	0.9399	0.9245
2.7181	225	1.0296	0.9400	0.9426	0.9293	0.9466	0.9345
3.0189	250	0.3606	0.9342	0.9373	0.9231	0.9417	0.9282
3.3212	275	1.2364	0.9381	0.9410	0.9273	0.9444	0.9312
3.6236	300	1.2507	0.9305	0.9340	0.9193	0.9385	0.9233
3.9259	325	0.3211	0.9500	0.9512	0.9417	0.9540	0.9460

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.0.1
Transformers: 4.41.2
PyTorch: 2.1.2+cu121
Accelerate: 0.31.0
Datasets: 2.19.1
Tokenizers: 0.19.1

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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning}, 
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@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}
}

Adi-0-0-Gupta
/

Embedding-v2

SentenceTransformer

Model Details

Model Description

Model Sources

Full Model Architecture

Usage

Direct Usage (Sentence Transformers)

Evaluation

Metrics

Information Retrieval

Information Retrieval

Information Retrieval

Information Retrieval

Information Retrieval

Training Details

Training Dataset

Unnamed Dataset

Training Hyperparameters

Non-Default Hyperparameters

All Hyperparameters

Training Logs

Framework Versions

Citation

BibTeX

Sentence Transformers

MatryoshkaLoss

MultipleNegativesRankingLoss

Evaluation results