This is a sentence-transformers model finetuned from intfloat/multilingual-e5-small. 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: intfloat/multilingual-e5-small
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': 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("srikarvar/multilingual-e5-small-triplet-final")
# Run inference
sentences = [
    'What is the capital city of Canada?',
    'What is the capital of Canada?',
    'What is the capital city of Australia?',
]
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

Triplet

Dataset: triplet-validation
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9836
dot_accuracy	0.0164
manhattan_accuracy	0.9836
euclidean_accuracy	0.9836
max_accuracy	0.9836

Training Details

Training Dataset

Unnamed Dataset

Size: 546 training samples
Columns: anchor, positive, and negative
Approximate statistics based on the first 1000 samples:
anchor positive negative
type string string string
details
min: 6 tokens
mean: 10.78 tokens
max: 22 tokens
min: 4 tokens
mean: 9.52 tokens
max: 19 tokens
min: 6 tokens
mean: 10.75 tokens
max: 22 tokens

Samples:

anchor	positive	negative
`What is the capital of Brazil?`	`Capital city of Brazil`	`What is the capital of Argentina?`
`How do I install Python on my computer?`	`How do I set up Python on my PC?`	`How do I uninstall Python on my computer?`
`How do I apply for a credit card?`	`How do I get a credit card?`	`How do I cancel a credit card?`

Loss: TripletLoss with these parameters:
json
1{ 2 "distance_metric": "TripletDistanceMetric.EUCLIDEAN", 3 "triplet_margin": 5 4}

Evaluation Dataset

Unnamed Dataset

Size: 61 evaluation samples
Columns: anchor, positive, and negative
Approximate statistics based on the first 1000 samples:
anchor positive negative
type string string string
details
min: 7 tokens
mean: 10.66 tokens
max: 16 tokens
min: 5 tokens
mean: 9.43 tokens
max: 14 tokens
min: 6 tokens
mean: 10.54 tokens
max: 17 tokens

Samples:

anchor	positive	negative
`How to create a podcast?`	`Steps to start a podcast`	`How to create a vlog?`
`How many states are there in the USA?`	`Total number of states in the United States`	`How many provinces are there in Canada?`
`What is the population of India?`	`How many people live in India?`	`What is the population of China?`

Loss: TripletLoss with these parameters:
json
1{ 2 "distance_metric": "TripletDistanceMetric.EUCLIDEAN", 3 "triplet_margin": 5 4}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
gradient_accumulation_steps: 2
learning_rate: 5e-06
weight_decay: 0.01
num_train_epochs: 12
lr_scheduler_type: cosine
warmup_steps: 50
load_best_model_at_end: True
optim: adamw_torch_fused

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
prediction_loss_only: True
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 2
eval_accumulation_steps: None
learning_rate: 5e-06
weight_decay: 0.01
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 12
max_steps: -1
lr_scheduler_type: cosine
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
warmup_steps: 50
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: 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: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	triplet-validation_max_accuracy
0.5714	10	4.9735	-	-
0.9714	17	-	4.9198	-
1.1429	20	4.9596	-	-
1.7143	30	4.9357	-	-
2.0	35	-	4.8494	-
2.2857	40	4.896	-	-
2.8571	50	4.8587	-	-
2.9714	52	-	4.7479	-
3.4286	60	4.8265	-	-
4.0	70	4.7706	4.6374	-
4.5714	80	4.7284	-	-
4.9714	87	-	4.5422	-
5.1429	90	4.6767	-	-
5.7143	100	4.653	-	-
6.0	105	-	4.4474	-
6.2857	110	4.6234	-	-
6.8571	120	4.5741	-	-
6.9714	122	-	4.3708	-
7.4286	130	4.5475	-	-
8.0	140	4.5206	4.3162	-
8.5714	150	4.517	-	-
8.9714	157	-	4.2891	-
9.1429	160	4.4587	-	-
9.7143	170	4.4879	-	-
10.0	175	-	4.2755	-
10.2857	180	4.4625	-	-
10.8571	190	4.489	-	-
10.9714	192	-	4.2716	-
11.4286	200	4.4693	-	-
11.6571	204	-	4.2713	0.9836

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.0.1
Transformers: 4.41.2
PyTorch: 2.1.2+cu121
Accelerate: 0.32.1
Datasets: 2.19.1
Tokenizers: 0.19.1

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",
}

TripletLoss

bibtex

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification}, 
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}

	anchor	positive	negative
type	string	string	string
details	min: 6 tokens mean: 10.78 tokens max: 22 tokens	min: 4 tokens mean: 9.52 tokens max: 19 tokens	min: 6 tokens mean: 10.75 tokens max: 22 tokens