magnifi/bge-small-en-v1.5-ft-orc-0806

SentenceTransformer based on BAAI/bge-small-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-small-en-v1.5. 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: BAAI/bge-small-en-v1.5
• 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': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'Which of my investments are projected to generate the most return?',
    '[{"get_portfolio(None)": "portfolio"}, {"get_expected_attribute(\'portfolio\',[\'returns\'])": "portfolio"}, {"sort(\'portfolio\',\'returns\',\'desc\')": "portfolio"}]',
    '[{"get_portfolio(None)": "portfolio"}, {"factor_contribution(\'portfolio\',\'<DATES>\',\'asset_class\',\'us equity\',\'returns\')": "portfolio"}]',
]
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

Information Retrieval

• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6644
cosine_accuracy@3	0.8288
cosine_accuracy@5	0.863
cosine_accuracy@10	0.9178
cosine_precision@1	0.6644
cosine_precision@3	0.2763
cosine_precision@5	0.1726
cosine_precision@10	0.0918
cosine_recall@1	0.0185
cosine_recall@3	0.023
cosine_recall@5	0.024
cosine_recall@10	0.0255
cosine_ndcg@10	0.1737
cosine_mrr@10	0.748
cosine_map@100	0.0209
dot_accuracy@1	0.6644
dot_accuracy@3	0.8288
dot_accuracy@5	0.863
dot_accuracy@10	0.9178
dot_precision@1	0.6644
dot_precision@3	0.2763
dot_precision@5	0.1726
dot_precision@10	0.0918
dot_recall@1	0.0185
dot_recall@3	0.023
dot_recall@5	0.024
dot_recall@10	0.0255
dot_ndcg@10	0.1737
dot_mrr@10	0.748
dot_map@100	0.0209

Training Details

Training Dataset

Unnamed Dataset

• Size: 723 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: 5 tokens mean: 11.8 tokens max: 26 tokens	min: 24 tokens mean: 84.41 tokens max: 194 tokens

• Samples:

  sentence_0	sentence_1
  what is my portfolio 3 year cagr?	[{"get_portfolio(None)": "portfolio"}, {"get_attribute('portfolio',['gains'],'')": "portfolio"}, {"sort('portfolio','gains','desc')": "portfolio"}]
  what is my 1 year rate of return	[{"get_portfolio(None)": "portfolio"}, {"get_attribute('portfolio',['gains'],'')": "portfolio"}, {"sort('portfolio','gains','desc')": "portfolio"}]
  show backtest of my performance this year?	[{"get_portfolio(None)": "portfolio"}, {"get_attribute('portfolio',['gains'],'')": "portfolio"}, {"sort('portfolio','gains','desc')": "portfolio"}]

• 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: 10
• per_device_eval_batch_size: 10
• num_train_epochs: 6
• 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: 10
• per_device_eval_batch_size: 10
• 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: 6
• 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
• 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: 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
• 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: round_robin

Training Logs

Epoch	Step	cosine_map@100
0.0274	2	0.0136
0.0548	4	0.0137
0.0822	6	0.0139
0.1096	8	0.0142
0.1370	10	0.0145
0.1644	12	0.0144
0.1918	14	0.0147
0.2192	16	0.0151
0.2466	18	0.0153
0.2740	20	0.0158
0.3014	22	0.0165
0.3288	24	0.0163
0.3562	26	0.0167
0.3836	28	0.0171
0.4110	30	0.0175
0.4384	32	0.0177
0.4658	34	0.0180
0.4932	36	0.0183
0.5205	38	0.0185
0.5479	40	0.0186
0.5753	42	0.0186
0.6027	44	0.0186
0.6301	46	0.0186
0.6575	48	0.0187
0.6849	50	0.0189
0.7123	52	0.0190
0.7397	54	0.0189
0.7671	56	0.0188
0.7945	58	0.0189
0.8219	60	0.0192
0.8493	62	0.0193
0.8767	64	0.0194
0.9041	66	0.0194
0.9315	68	0.0197
0.9589	70	0.0200
0.9863	72	0.0201
1.0	73	0.0202
1.0137	74	0.0203
1.0411	76	0.0202
1.0685	78	0.0203
1.0959	80	0.0205
1.1233	82	0.0207
1.1507	84	0.0207
1.1781	86	0.0206
1.2055	88	0.0205
1.2329	90	0.0205
1.2603	92	0.0205
1.2877	94	0.0204
1.3151	96	0.0204
1.3425	98	0.0205
1.3699	100	0.0205
1.3973	102	0.0205
1.4247	104	0.0205
1.4521	106	0.0204
1.4795	108	0.0205
1.5068	110	0.0208
1.5342	112	0.0206
1.5616	114	0.0205
1.5890	116	0.0206
1.6164	118	0.0205
1.6438	120	0.0205
1.6712	122	0.0205
1.6986	124	0.0207
1.7260	126	0.0207
1.7534	128	0.0207
1.7808	130	0.0205
1.8082	132	0.0206
1.8356	134	0.0208
1.8630	136	0.0206
1.8904	138	0.0206
1.9178	140	0.0206
1.9452	142	0.0205
1.9726	144	0.0206
2.0	146	0.0207
2.0274	148	0.0209

Framework Versions

• Python: 3.10.9
• Sentence Transformers: 3.0.1
• Transformers: 4.44.0
• PyTorch: 2.4.0+cu121
• Accelerate: 0.33.0
• 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}
}