This is a Cross Encoder model finetuned from cross-encoder/mmarco-mMiniLMv2-L12-H384-v1 on the arabic-ecom-data dataset using the sentence-transformers library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.

Model Details

Model Description

Model Type: Cross Encoder
Base model: cross-encoder/mmarco-mMiniLMv2-L12-H384-v1
Maximum Sequence Length: 128 tokens
Number of Output Labels: 1 label
Supported Modality: Text
Training Dataset:
- arabic-ecom-data

Model Sources

Documentation: Sentence Transformers Documentation
Documentation: Cross Encoder Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Cross Encoders on Hugging Face

Full Model Architecture

CrossEncoder(
  (0): Transformer({'transformer_task': 'sequence-classification', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'logits'}}, 'module_output_name': 'scores', 'architecture': 'XLMRobertaForSequenceClassification'})
)

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("leafxyz/arabic-ecom-cross-encoder-v3")
# Get scores for pairs of inputs
pairs = [
    ['مناديل مطبخ', 'صابون اواني جودي - 960 مل (الليمون الاخضر)'],
    ['جبنة هابي كاو', 'هابي كاو جبنة كريمى - 150 غ'],
    ['كريم تايغر للشعر', 'كريم ازالة شعر - Page Vine'],
    ['لانشون حلواني', 'لانشون حلواني دجاج - 250 غ'],
    ['صابون جودي 2.32', 'صابون اواني جودي برائحة الليمون الاخضر - 2.32 ل'],
]
scores = model.predict(pairs)
print(scores)
# [-5.0312  0.2981 -1.2588  0.6904  0.7002]

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'مناديل مطبخ',
    [
        'صابون اواني جودي - 960 مل (الليمون الاخضر)',
        'هابي كاو جبنة كريمى - 150 غ',
        'كريم ازالة شعر - Page Vine',
        'لانشون حلواني دجاج - 250 غ',
        'صابون اواني جودي برائحة الليمون الاخضر - 2.32 ل',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Training Details

Training Dataset

arabic-ecom-data

Dataset: arabic-ecom-data
Size: 246,013 training samples
Columns: sentence1, sentence2, and label
Approximate statistics based on the first 1000 samples:
sentence1 sentence2 label
type string string float
details
min: 3 tokens
mean: 7.77 tokens
max: 20 tokens
min: 4 tokens
mean: 13.06 tokens
max: 27 tokens
min: 0.0
mean: 0.49
max: 1.0

Samples:

sentence1	sentence2	label
`فلوتس أصبع`	`كيت كات شوكلاتة 4 اصابع 36.5 جم`	`0.0`
`بخور عود ند شيخ العرب`	`بخور العود- اصل العود`	`0.0`
`احمر شفاه Rhode`	`احمر شفاه - Water Lip Matte`	`0.0`

Loss: BinaryCrossEntropyLoss with these parameters:
json
1{ 2 "activation_fn": "torch.nn.modules.linear.Identity", 3 "pos_weight": null 4}

Evaluation Dataset

arabic-ecom-data

Dataset: arabic-ecom-data
Size: 5,021 evaluation samples
Columns: sentence1, sentence2, and label
Approximate statistics based on the first 1000 samples:
sentence1 sentence2 label
type string string float
details
min: 3 tokens
mean: 7.86 tokens
max: 16 tokens
min: 4 tokens
mean: 13.19 tokens
max: 33 tokens
min: 0.0
mean: 0.48
max: 1.0

Samples:

sentence1	sentence2	label
`مناديل مطبخ`	`صابون اواني جودي - 960 مل (الليمون الاخضر)`	`0.0`
`جبنة هابي كاو`	`هابي كاو جبنة كريمى - 150 غ`	`1.0`
`كريم تايغر للشعر`	`كريم ازالة شعر - Page Vine`	`0.0`

Loss: BinaryCrossEntropyLoss with these parameters:
json
1{ 2 "activation_fn": "torch.nn.modules.linear.Identity", 3 "pos_weight": null 4}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 32
per_device_eval_batch_size: 32
learning_rate: 2e-05
num_train_epochs: 2
warmup_steps: 0.1
fp16: True

All Hyperparameters

Click to expand

do_predict: False
prediction_loss_only: True
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
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: 2
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: None
warmup_ratio: None
warmup_steps: 0.1
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
enable_jit_checkpoint: False
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
use_cpu: False
seed: 42
data_seed: None
bf16: False
fp16: True
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: -1
ddp_backend: None
debug: []
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_prefetch_factor: None
disable_tqdm: False
remove_unused_columns: True
label_names: None
load_best_model_at_end: False
ignore_data_skip: False
fsdp: []
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
parallelism_config: None
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch_fused
optim_args: None
group_by_length: False
length_column_name: length
project: huggingface
trackio_space_id: trackio
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
push_to_hub: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: None
hub_always_push: False
hub_revision: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_for_metrics: []
eval_do_concat_batches: True
auto_find_batch_size: False
full_determinism: False
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
include_num_input_tokens_seen: no
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: True
use_cache: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss	Validation Loss
0.0130	100	0.8919	-
0.0260	200	0.6599	-
0.0390	300	0.5613	-
0.0520	400	0.5168	-
0.0650	500	0.5278	0.4916
0.0780	600	0.5182	-
0.0911	700	0.4833	-
0.1041	800	0.4863	-
0.1171	900	0.5011	-
0.1301	1000	0.4740	0.4477
0.1431	1100	0.4480	-
0.1561	1200	0.4536	-
0.1691	1300	0.4604	-
0.1821	1400	0.4704	-
0.1951	1500	0.4514	0.4282
0.2081	1600	0.4358	-
0.2211	1700	0.4472	-
0.2341	1800	0.4382	-
0.2471	1900	0.4524	-
0.2601	2000	0.4368	0.4112
0.2732	2100	0.4272	-
0.2862	2200	0.4280	-
0.2992	2300	0.4276	-
0.3122	2400	0.4067	-
0.3252	2500	0.4260	0.4026
0.3382	2600	0.4321	-
0.3512	2700	0.4333	-
0.3642	2800	0.4246	-
0.3772	2900	0.4304	-
0.3902	3000	0.4237	0.3938
0.4032	3100	0.4181	-
0.4162	3200	0.4224	-
0.4292	3300	0.4096	-
0.4422	3400	0.4069	-
0.4553	3500	0.4045	0.3963
0.4683	3600	0.4164	-
0.4813	3700	0.3996	-
0.4943	3800	0.4053	-
0.5073	3900	0.3853	-
0.5203	4000	0.4035	0.3818
0.5333	4100	0.4043	-
0.5463	4200	0.3914	-
0.5593	4300	0.4022	-
0.5723	4400	0.3949	-
0.5853	4500	0.4094	0.3821
0.5983	4600	0.3782	-
0.6113	4700	0.3908	-
0.6243	4800	0.3944	-
0.6374	4900	0.4112	-
0.6504	5000	0.4077	0.3676
0.6634	5100	0.4034	-
0.6764	5200	0.3958	-
0.6894	5300	0.3988	-
0.7024	5400	0.3835	-
0.7154	5500	0.4065	0.3680
0.7284	5600	0.3910	-
0.7414	5700	0.3959	-
0.7544	5800	0.4005	-
0.7674	5900	0.3967	-
0.7804	6000	0.3947	0.3734
0.7934	6100	0.3916	-
0.8065	6200	0.4023	-
0.8195	6300	0.3869	-
0.8325	6400	0.3821	-
0.8455	6500	0.3845	0.3716
0.8585	6600	0.3637	-
0.8715	6700	0.3828	-
0.8845	6800	0.3703	-
0.8975	6900	0.3962	-
0.9105	7000	0.3880	0.3592
0.9235	7100	0.3846	-
0.9365	7200	0.3722	-
0.9495	7300	0.3946	-
0.9625	7400	0.3779	-
0.9755	7500	0.3957	0.3550
0.9886	7600	0.3763	-
1.0016	7700	0.3732	-
1.0146	7800	0.3763	-
1.0276	7900	0.3713	-
1.0406	8000	0.3594	0.3597
1.0536	8100	0.3510	-
1.0666	8200	0.3738	-
1.0796	8300	0.3554	-
1.0926	8400	0.3524	-
1.1056	8500	0.3507	0.3577
1.1186	8600	0.3483	-
1.1316	8700	0.3692	-
1.1446	8800	0.3676	-
1.1576	8900	0.3484	-
1.1707	9000	0.3859	0.3502
1.1837	9100	0.3590	-
1.1967	9200	0.3746	-
1.2097	9300	0.3559	-
1.2227	9400	0.3631	-
1.2357	9500	0.3500	0.3685
1.2487	9600	0.3496	-
1.2617	9700	0.3803	-
1.2747	9800	0.3442	-
1.2877	9900	0.3503	-
1.3007	10000	0.3636	0.3504
1.3137	10100	0.3479	-
1.3267	10200	0.3768	-
1.3398	10300	0.3501	-
1.3528	10400	0.3563	-
1.3658	10500	0.3551	0.3515
1.3788	10600	0.3645	-
1.3918	10700	0.3466	-
1.4048	10800	0.3622	-
1.4178	10900	0.3535	-
1.4308	11000	0.3708	0.3452
1.4438	11100	0.3484	-
1.4568	11200	0.3593	-
1.4698	11300	0.3554	-
1.4828	11400	0.3362	-
1.4958	11500	0.3707	0.3458
1.5088	11600	0.3559	-
1.5219	11700	0.3501	-
1.5349	11800	0.3771	-
1.5479	11900	0.3586	-
1.5609	12000	0.3462	0.3478
1.5739	12100	0.3448	-
1.5869	12200	0.3516	-
1.5999	12300	0.3582	-
1.6129	12400	0.3621	-
1.6259	12500	0.3724	0.3436
1.6389	12600	0.3598	-
1.6519	12700	0.3616	-
1.6649	12800	0.3537	-
1.6779	12900	0.3462	-
1.6909	13000	0.3675	0.3443
1.7040	13100	0.3506	-
1.7170	13200	0.3389	-
1.7300	13300	0.3454	-
1.7430	13400	0.3588	-
1.7560	13500	0.3521	0.3427
1.7690	13600	0.3462	-
1.7820	13700	0.3513	-
1.7950	13800	0.3484	-
1.8080	13900	0.3522	-
1.8210	14000	0.3426	0.3447
1.8340	14100	0.3497	-
1.8470	14200	0.3464	-
1.8600	14300	0.3427	-
1.8730	14400	0.3422	-
1.8861	14500	0.3398	0.3447
1.8991	14600	0.3487	-
1.9121	14700	0.3608	-
1.9251	14800	0.3515	-
1.9381	14900	0.3456	-
1.9511	15000	0.3499	0.3445
1.9641	15100	0.3404	-
1.9771	15200	0.3482	-
1.9901	15300	0.3464	-

Training Time

Training: 21.4 minutes

Framework Versions

Python: 3.12.13
Sentence Transformers: 5.4.1
Transformers: 5.0.0
PyTorch: 2.10.0+cu128
Accelerate: 1.13.0
Datasets: 5.0.0
Tokenizers: 0.22.2

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",
}

	sentence1	sentence2	label
type	string	string	float
details	min: 3 tokens mean: 7.77 tokens max: 20 tokens	min: 4 tokens mean: 13.06 tokens max: 27 tokens	min: 0.0 mean: 0.49 max: 1.0