SentenceTransformer based on FacebookAI/xlm-roberta-base

This is a sentence-transformers model finetuned from FacebookAI/xlm-roberta-base on the en-sa dataset. 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: FacebookAI/xlm-roberta-base
Maximum Sequence Length: 128 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- en-sa

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': 128, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (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("saikasyap/samskritam-xlm-roberata-base")
# Run inference
sentences = [
    'अनन्तरं Microbiology इति टङ्कनं करोमि ।',
    'Then I will type Microbiology.',
    'The comic image basically consists of three parts.',
]
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]

Training Details

Training Dataset

en-sa

Dataset: en-sa
Size: 140,359 training samples
Columns: non_english, english, and label
Approximate statistics based on the first 1000 samples:
non_english english label
type string string list
details
min: 26 tokens
mean: 42.11 tokens
max: 128 tokens

min: 19 tokens
mean: 50.63 tokens
max: 128 tokens

size: 768 elements

	non_english	english	label
type	string	string	list
details	min: 26 tokens mean: 42.11 tokens max: 128 tokens	min: 19 tokens mean: 50.63 tokens max: 128 tokens	size: 768 elements

Samples:

non_english	english	label
`ॐ तपः स्वाध्यायनिरतं तपस्वी वाग्विदां वरम्। नारदं परिपप्रच्छ वाल्मीकिर्मुनिपुङ्गवम्॥`	`The ascetic Vālmīki asked Nārada, the best of sages and foremost of those conversant with words, ever engaged in austerities and Vedic studies.`	`[0.15034635365009308, 0.35359007120132446, -0.3348075747489929, 0.15415771305561066, 0.020571526139974594, ...]`
`कोन्वस्मिन् साम्प्रतं लोके गुणवान् कश्च वीर्यवान्। धर्मज्ञश्च कृतज्ञश्च सत्यवाक्यो दृढत्नतः॥`	`Who at present in this world is like crowned with qualities, and with prowess, knowing duty, and grateful, and truthful, and firm in vow.`	`[-0.46556514501571655, 0.4740210175514221, -0.2033461034297943, -1.6129034757614136, -0.016881834715604782, ...]`
`चारित्रेण च को युक्तः सर्वभूतेषु को हितः। विद्वान् कः कः समर्थश्च कश्चैकप्रियदर्शनः॥`	`Who is qualified by virtue of his character, and who is engaged in the welfare of all creatures? Who is learned and capable. Who alone is ever lovely to behold?`	`[-0.09693514555692673, 0.4206468462944031, -0.3034357726573944, -1.2955875396728516, 0.3836270868778229, ...]`

Loss: MSELoss

Evaluation Dataset

en-sa

Dataset: en-sa
Size: 1,000 evaluation samples
Columns: non_english, english, and label
Approximate statistics based on the first 1000 samples:
non_english english label
type string string list
details
min: 4 tokens
mean: 27.87 tokens
max: 91 tokens

min: 4 tokens
mean: 21.35 tokens
max: 68 tokens

size: 768 elements

	non_english	english	label
type	string	string	list
details	min: 4 tokens mean: 27.87 tokens max: 91 tokens	min: 4 tokens mean: 21.35 tokens max: 68 tokens	size: 768 elements

Samples:

non_english	english	label
`तथा दिमागी रूप से तंदुरुस्त हों ।`	`And also be mentally fit.`	`[0.2053176611661911, -0.15136581659317017, -0.1492331326007843, -0.13915303349494934, -0.08056919276714325, ...]`
`अपरञ्च युष्माकम् आनन्दो यत् सम्पूर्णो भवेद् तदर्थं वयम् एतानि लिखामः।`	`"""And these things write we unto you, that your joy may be full."""`	`[0.0013895286247134209, 0.09506042301654816, -0.3513864576816559, -0.6496815085411072, 0.7649527192115784, ...]`
`पञ्च व्यञ्जनानां तेषां च एकम्-एकम स्वास्थ्यसम्बन्धीकार्याणां च सूचीं निर्मातु।`	`List five spices and one health benefits of each.`	`[0.37307825684547424, 0.8675527572631836, 0.6388981342315674, -0.27114301919937134, -0.30143851041793823, ...]`

Loss: MSELoss

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 64
per_device_eval_batch_size: 64
learning_rate: 2e-05
num_train_epochs: 5
warmup_ratio: 0.1
fp16: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 64
per_device_eval_batch_size: 64
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: 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: 5
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: 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: False
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

Training Logs

Epoch	Step	Training Loss
0.1821	100	0.5076
0.3643	200	0.233
0.5464	300	0.2172
0.7286	400	0.2033
0.9107	500	0.1943
1.0929	600	0.1878
1.2750	700	0.1813
1.4572	800	0.1754
1.6393	900	0.1721
1.8215	1000	0.1688
2.0036	1100	0.1664
2.1858	1200	0.1632
2.3679	1300	0.1606
2.5501	1400	0.1588
2.7322	1500	0.1566
2.9144	1600	0.1558
3.0965	1700	0.154
3.2787	1800	0.1525
3.4608	1900	0.1508
3.6430	2000	0.15
3.8251	2100	0.1493
4.0073	2200	0.149
4.1894	2300	0.1479
4.3716	2400	0.1471
4.5537	2500	0.1466
4.7359	2600	0.1461
4.9180	2700	0.1466

Framework Versions

Python: 3.10.17
Sentence Transformers: 4.1.0
Transformers: 4.46.3
PyTorch: 2.2.0+cu121
Accelerate: 1.1.1
Datasets: 2.18.0
Tokenizers: 0.20.3

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",
}

MSELoss

@inproceedings{reimers-2020-multilingual-sentence-bert,
    title = "Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2020",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/2004.09813",
}