hyrinmansoor/changAI-nomic-embed-text-v1.5-finetuned

SentenceTransformer based on nomic-ai/nomic-embed-text-v1.5

This is a sentence-transformers model finetuned from nomic-ai/nomic-embed-text-v1.5. 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: nomic-ai/nomic-embed-text-v1.5
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 768 dimensions
• 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': 8192, 'do_lower_case': False, 'architecture': 'NomicBertModel'})
  (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("sentence_transformers_model_id")
# Run inference
sentences = [
    'search_query: quality inspection rejected but material received',
    'search_document: [FIELD] quality_inspection | [TABLE] tabPurchase Receipt Item',
    'search_document: [FIELD] delivery_date | [TABLE] tabSales Order',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.7248, 0.0996],
#         [0.7248, 1.0000, 0.2103],
#         [0.0996, 0.2103, 1.0000]])

Evaluation

Metrics

Information Retrieval

• Dataset: erp-val
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5228
cosine_accuracy@5	0.7842
cosine_accuracy@12	0.8759
cosine_precision@12	0.1248
cosine_recall@12	0.8324
cosine_ndcg@10	0.6569
cosine_mrr@10	0.6305
cosine_map@10	0.5873
cosine_map@12	0.5905

Training Details

Training Dataset

Unnamed Dataset

• Size: 128,055 training samples
• Columns: sentence_0 and sentence_1
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1
  type	string	string
  details	min: 9 tokens mean: 15.81 tokens max: 34 tokens	min: 11 tokens mean: 19.35 tokens max: 49 tokens

• Samples:

  sentence_0	sentence_1
  search_query: where's the list of folks on leave?	search_document: [FIELD] status | [TABLE] tabLeave Application
  search_query: 2 barcodes same	search_document: [FIELD] barcode | [TABLE] tabItem Barcode
  search_query: customers who haven’t bought last 12 mo	search_document: [FIELD] posting_date | [TABLE] tabSales Invoice

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 40.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 48
• per_device_eval_batch_size: 48
• num_train_epochs: 4
• fp16: True
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 48
• per_device_eval_batch_size: 48
• 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: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1
• num_train_epochs: 4
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• 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
• 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}
• parallelism_config: 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
• 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
• use_legacy_prediction_loop: False
• 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_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
• include_tokens_per_second: False
• 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
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	erp-val_cosine_ndcg@10
-1	-1	-	0.4109
0.0562	150	-	0.5393
0.1124	300	-	0.5979
0.1687	450	-	0.6070
0.1874	500	2.226	-
0.2249	600	-	0.6149
0.2811	750	-	0.6232
0.3373	900	-	0.6207
0.3748	1000	1.557	-
0.3936	1050	-	0.6333
0.4498	1200	-	0.6410
0.5060	1350	-	0.6392
0.5622	1500	1.3475	0.6439
0.6184	1650	-	0.6402
0.6747	1800	-	0.6405
0.7309	1950	-	0.6465
0.7496	2000	1.249	-
0.7871	2100	-	0.6439
0.8433	2250	-	0.6465
0.8996	2400	-	0.6377
0.9370	2500	1.1999	-
0.9558	2550	-	0.6418
1.0	2668	-	0.6400
1.0120	2700	-	0.6414
1.0682	2850	-	0.6565
1.1244	3000	1.1002	0.6475
1.1807	3150	-	0.6440
1.2369	3300	-	0.6449
1.2931	3450	-	0.6540
1.3118	3500	1.0786	-
1.3493	3600	-	0.6492
1.4055	3750	-	0.6521
1.4618	3900	-	0.6511
1.4993	4000	1.0513	-
1.5180	4050	-	0.6595
1.5742	4200	-	0.6512
1.6304	4350	-	0.6456
1.6867	4500	1.0218	0.6508
1.7429	4650	-	0.6500
1.7991	4800	-	0.6405
1.8553	4950	-	0.6612
1.8741	5000	1.0101	-
1.9115	5100	-	0.6462
1.9678	5250	-	0.6571
2.0	5336	-	0.6616
2.0240	5400	-	0.6654
2.0615	5500	0.977	-
2.0802	5550	-	0.6554
2.1364	5700	-	0.6536
2.1927	5850	-	0.6614
2.2489	6000	0.9208	0.6629
2.3051	6150	-	0.6576
2.3613	6300	-	0.6503
2.4175	6450	-	0.6630
2.4363	6500	0.935	-
2.4738	6600	-	0.6509
2.5300	6750	-	0.6570
2.5862	6900	-	0.6485
2.6237	7000	0.9224	-
2.6424	7050	-	0.6508
2.6987	7200	-	0.6587
2.7549	7350	-	0.6578
2.8111	7500	0.9068	0.6563
2.8673	7650	-	0.6565
2.9235	7800	-	0.6612
2.9798	7950	-	0.6608
2.9985	8000	0.9139	-
3.0	8004	-	0.6588
3.0360	8100	-	0.6579
3.0922	8250	-	0.6526
3.1484	8400	-	0.6567
3.1859	8500	0.8468	-
3.2046	8550	-	0.6553
3.2609	8700	-	0.6562
3.3171	8850	-	0.6570
3.3733	9000	0.8638	0.6551
3.4295	9150	-	0.6551
3.4858	9300	-	0.6551
3.5420	9450	-	0.6564
3.5607	9500	0.8388	-
3.5982	9600	-	0.6563
3.6544	9750	-	0.6568
3.7106	9900	-	0.6564
3.7481	10000	0.8544	-
3.7669	10050	-	0.6568
3.8231	10200	-	0.6561
3.8793	10350	-	0.6567
3.9355	10500	0.8349	0.6569

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.1.2
• Transformers: 4.57.2
• PyTorch: 2.9.0+cu126
• Accelerate: 1.12.0
• Datasets: 4.0.0
• Tokenizers: 0.22.1

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}
}