aidal/persian-sentence-transformer-product-classification

SentenceTransformer based on ahdsoft/persian-sentence-transformer-news-wiki-pairs-v3

This is a sentence-transformers model finetuned from ahdsoft/persian-sentence-transformer-news-wiki-pairs-v3. It maps sentences & paragraphs to a 1024-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: ahdsoft/persian-sentence-transformer-news-wiki-pairs-v3
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 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: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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("aidal/persian-sentence-transformer-product-classification")
# Run inference
sentences = [
    'بازی آموزشی مدل جورچین ایران کد K-5',
    'اسباب بازی، کودک و نوزاد',
    'زیبایی و سلامت',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Semantic Similarity

• Dataset: embedding-similarity-eval
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	nan
spearman_cosine	nan
pearson_manhattan	nan
spearman_manhattan	nan
pearson_euclidean	nan
spearman_euclidean	nan
pearson_dot	nan
spearman_dot	nan
pearson_max	nan
spearman_max	nan

Training Details

Training Dataset

Unnamed Dataset

• Size: 804,708 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 6 tokens mean: 16.15 tokens max: 41 tokens	min: 3 tokens mean: 6.42 tokens max: 11 tokens

• Samples:

  anchor	positive
  مربا زرشک مارجان - 270 گرم	محصولات بومی و محلی
  دفتر یادداشت بادکنک آبی طرح انیمه مدل Attack on titan مجموعه 2 عددی	کتاب، لوازم تحریر و هنر
  چای ساز کاراجا مدل Cay Sever	لوازم خانگی برقی

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 89,413 evaluation samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 6 tokens mean: 16.02 tokens max: 44 tokens	min: 3 tokens mean: 6.38 tokens max: 11 tokens

• Samples:

  anchor	positive
  لامپ ال ای دی 6 وات لداستار مدل شعله ای پایه E27 بسته 3 عددی	خانه و آشپزخانه
  زیرانداز تعویض نوزاد مدل هپی ویکند	اسباب بازی، کودک و نوزاد
  تابلو نوری کاکتی مدل عاشقانه طرح اسم شهسوار کد TA14352	خانه و آشپزخانه

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• learning_rate: 2e-05
• num_train_epochs: 1
• warmup_ratio: 0.1
• log_level: debug
• fp16: True
• load_best_model_at_end: 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: 32
• per_device_eval_batch_size: 32
• 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: 1
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 0
• log_level: debug
• 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: 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
• 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
• eval_use_gather_object: False
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss	loss	embedding-similarity-eval_spearman_max
0.0040	100	2.7527	-	-
0.0080	200	2.0773	-	-
0.0119	300	1.764	-	-
0.0159	400	1.5861	-	-
0.0199	500	1.5138	-	-
0.0239	600	1.4307	-	-
0.0278	700	1.3923	-	-
0.0318	800	1.3251	-	-
0.0358	900	1.3023	-	-
0.0398	1000	1.2929	-	-
0.0437	1100	1.2764	-	-
0.0477	1200	1.2728	-	-
0.0517	1300	1.2262	-	-
0.0557	1400	1.2456	-	-
0.0596	1500	1.2052	-	-
0.0636	1600	1.1912	-	-
0.0676	1700	1.2077	-	-
0.0716	1800	1.2196	-	-
0.0756	1900	1.1603	-	-
0.0795	2000	1.1706	-	-
0.0835	2100	1.2001	-	-
0.0875	2200	1.1822	-	-
0.0915	2300	1.1703	-	-
0.0954	2400	1.204	-	-
0.0994	2500	1.1863	1.1333	nan
0.1034	2600	1.1567	-	-
0.1074	2700	1.1876	-	-
0.1113	2800	1.1553	-	-
0.1153	2900	1.1332	-	-
0.1193	3000	1.1426	-	-
0.1233	3100	1.1476	-	-
0.1272	3200	1.1482	-	-
0.1312	3300	1.1343	-	-
0.1352	3400	1.1572	-	-
0.1392	3500	1.1018	-	-
0.1432	3600	1.1175	-	-
0.1471	3700	1.1024	-	-
0.1511	3800	1.1308	-	-
0.1551	3900	1.1386	-	-
0.1591	4000	1.1103	-	-
0.1630	4100	1.1472	-	-
0.1670	4200	1.1079	-	-
0.1710	4300	1.1199	-	-
0.1750	4400	1.1306	-	-
0.1789	4500	1.0975	-	-
0.1829	4600	1.1285	-	-
0.1869	4700	1.121	-	-
0.1909	4800	1.1099	-	-
0.1948	4900	1.0913	-	-
0.1988	5000	1.0631	1.0980	nan
0.2028	5100	1.1336	-	-
0.2068	5200	1.1055	-	-
0.2108	5300	1.0987	-	-
0.2147	5400	1.1078	-	-
0.2187	5500	1.0749	-	-
0.2227	5600	1.1016	-	-
0.2267	5700	1.0768	-	-
0.2306	5800	1.0954	-	-
0.2346	5900	1.0975	-	-
0.2386	6000	1.0638	-	-
0.2426	6100	1.0751	-	-
0.2465	6200	1.0675	-	-
0.2505	6300	1.0513	-	-
0.2545	6400	1.0808	-	-
0.2585	6500	1.0863	-	-
0.2624	6600	1.0681	-	-
0.2664	6700	1.0813	-	-
0.2704	6800	1.077	-	-
0.2744	6900	1.0811	-	-
0.2784	7000	1.0543	-	-
0.2823	7100	1.0677	-	-
0.2863	7200	1.0691	-	-
0.2903	7300	1.0597	-	-
0.2943	7400	1.0538	-	-
0.2982	7500	1.0853	1.0658	nan
0.3022	7600	1.0831	-	-
0.3062	7700	1.0565	-	-
0.3102	7800	1.0667	-	-
0.3141	7900	1.0839	-	-
0.3181	8000	1.0742	-	-
0.3221	8100	1.0543	-	-
0.3261	8200	1.0539	-	-
0.3300	8300	1.07	-	-
0.3340	8400	1.0556	-	-
0.3380	8500	1.0715	-	-
0.3420	8600	1.0468	-	-
0.3460	8700	1.0477	-	-
0.3499	8800	1.0401	-	-
0.3539	8900	1.1047	-	-
0.3579	9000	1.0345	-	-
0.3619	9100	1.0677	-	-
0.3658	9200	1.0705	-	-
0.3698	9300	1.0624	-	-
0.3738	9400	1.0528	-	-
0.3778	9500	1.0455	-	-
0.3817	9600	1.0555	-	-
0.3857	9700	1.0338	-	-
0.3897	9800	1.0624	-	-
0.3937	9900	1.0645	-	-
0.3976	10000	1.0622	1.0430	nan
0.4016	10100	1.0523	-	-
0.4056	10200	1.0697	-	-
0.4096	10300	1.0733	-	-
0.4136	10400	1.0415	-	-
0.4175	10500	1.0644	-	-
0.4215	10600	1.0404	-	-
0.4255	10700	1.026	-	-
0.4295	10800	1.0408	-	-
0.4334	10900	1.0602	-	-
0.4374	11000	1.0538	-	-
0.4414	11100	1.0396	-	-
0.4454	11200	1.0852	-	-
0.4493	11300	1.0412	-	-
0.4533	11400	1.0249	-	-
0.4573	11500	1.024	-	-
0.4613	11600	1.0494	-	-
0.4652	11700	1.0461	-	-
0.4692	11800	1.027	-	-
0.4732	11900	1.0802	-	-
0.4772	12000	1.0402	-	-
0.4812	12100	1.026	-	-
0.4851	12200	1.0565	-	-
0.4891	12300	1.0416	-	-
0.4931	12400	1.0452	-	-
0.4971	12500	1.0425	1.0376	nan
0.5010	12600	1.0319	-	-
0.5050	12700	1.0422	-	-
0.5090	12800	1.0261	-	-
0.5130	12900	1.0498	-	-
0.5169	13000	1.0189	-	-
0.5209	13100	1.0309	-	-
0.5249	13200	1.0509	-	-
0.5289	13300	1.0524	-	-
0.5328	13400	1.0516	-	-
0.5368	13500	1.0104	-	-
0.5408	13600	1.0394	-	-
0.5448	13700	1.0473	-	-
0.5488	13800	1.0151	-	-
0.5527	13900	1.0379	-	-
0.5567	14000	1.0556	-	-
0.5607	14100	1.0465	-	-
0.5647	14200	1.046	-	-
0.5686	14300	1.0211	-	-
0.5726	14400	1.0234	-	-
0.5766	14500	1.0215	-	-
0.5806	14600	1.0445	-	-
0.5845	14700	1.0229	-	-
0.5885	14800	1.0383	-	-
0.5925	14900	1.0491	-	-
0.5965	15000	1.0425	1.0303	nan
0.6004	15100	1.052	-	-
0.6044	15200	1.0281	-	-
0.6084	15300	1.0288	-	-
0.6124	15400	1.0096	-	-
0.6164	15500	1.0447	-	-
0.6203	15600	1.038	-	-
0.6243	15700	1.0061	-	-
0.6283	15800	1.0255	-	-
0.6323	15900	1.0246	-	-
0.6362	16000	1.0255	-	-
0.6402	16100	1.0271	-	-
0.6442	16200	1.0163	-	-
0.6482	16300	1.0381	-	-
0.6521	16400	1.0333	-	-
0.6561	16500	1.0161	-	-
0.6601	16600	1.03	-	-
0.6641	16700	1.0299	-	-
0.6680	16800	1.0191	-	-
0.6720	16900	1.0268	-	-
0.6760	17000	1.0177	-	-
0.6800	17100	1.0157	-	-
0.6840	17200	1.0382	-	-
0.6879	17300	1.0306	-	-
0.6919	17400	1.0231	-	-
0.6959	17500	1.0456	1.0231	nan
0.6999	17600	0.9993	-	-
0.7038	17700	1.0212	-	-
0.7078	17800	1.0114	-	-
0.7118	17900	1.0169	-	-
0.7158	18000	1.0115	-	-
0.7197	18100	1.019	-	-
0.7237	18200	1.016	-	-
0.7277	18300	1.0252	-	-
0.7317	18400	1.0374	-	-
0.7356	18500	1.0147	-	-
0.7396	18600	1.0302	-	-
0.7436	18700	1.0203	-	-
0.7476	18800	1.0395	-	-
0.7516	18900	1.0486	-	-
0.7555	19000	1.0321	-	-
0.7595	19100	1.0463	-	-
0.7635	19200	1.0124	-	-
0.7675	19300	1.0026	-	-
0.7714	19400	1.0474	-	-
0.7754	19500	1.0314	-	-
0.7794	19600	1.0183	-	-
0.7834	19700	1.0067	-	-
0.7873	19800	1.0179	-	-
0.7913	19900	1.0388	-	-
0.7953	20000	1.0063	1.0157	nan
0.7993	20100	1.0175	-	-
0.8032	20200	1.0349	-	-
0.8072	20300	1.0125	-	-
0.8112	20400	0.9982	-	-
0.8152	20500	1.0428	-	-
0.8192	20600	1.0526	-	-
0.8231	20700	1.0424	-	-
0.8271	20800	1.008	-	-
0.8311	20900	1.0186	-	-
0.8351	21000	1.0256	-	-
0.8390	21100	1.0125	-	-
0.8430	21200	1.0286	-	-
0.8470	21300	1.0358	-	-
0.8510	21400	1.0189	-	-
0.8549	21500	0.9861	-	-
0.8589	21600	0.9934	-	-
0.8629	21700	1.0211	-	-
0.8669	21800	1.0221	-	-
0.8708	21900	1.0302	-	-
0.8748	22000	1.0145	-	-
0.8788	22100	1.0027	-	-
0.8828	22200	1.0084	-	-
0.8868	22300	1.0334	-	-
0.8907	22400	1.0025	-	-
0.8947	22500	1.0175	1.0102	nan
0.8987	22600	1.0	-	-
0.9027	22700	1.0268	-	-
0.9066	22800	0.9795	-	-
0.9106	22900	1.0071	-	-
0.9146	23000	1.0141	-	-
0.9186	23100	1.006	-	-
0.9225	23200	1.0327	-	-
0.9265	23300	1.0016	-	-
0.9305	23400	1.0313	-	-
0.9345	23500	1.021	-	-
0.9384	23600	1.0217	-	-
0.9424	23700	1.0191	-	-
0.9464	23800	1.0238	-	-
0.9504	23900	1.0469	-	-
0.9544	24000	1.0338	-	-
0.9583	24100	1.0043	-	-
0.9623	24200	1.0054	-	-
0.9663	24300	1.0264	-	-
0.9703	24400	1.024	-	-
0.9742	24500	1.0172	-	-
0.9782	24600	1.0127	-	-
0.9822	24700	1.013	-	-
0.9862	24800	1.0135	-	-
0.9901	24900	1.0145	-	-
0.9941	25000	1.0184	1.0082	nan
0.9981	25100	1.0305	-	-

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.43.3
• PyTorch: 2.2.2+cu121
• Accelerate: 0.32.1
• Datasets: 2.20.0
• 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",
}

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}
}