Fingu-M-v2 / README.md
FINGU-AI's picture
Upload folder using huggingface_hub
2a44ea1 verified
metadata
base_model: dunzhang/stella_en_1.5B_v5
datasets: []
language: []
library_name: sentence-transformers
metrics:
  - cosine_accuracy@1
  - cosine_accuracy@3
  - cosine_accuracy@5
  - cosine_accuracy@10
  - cosine_precision@1
  - cosine_precision@3
  - cosine_precision@5
  - cosine_precision@10
  - cosine_recall@1
  - cosine_recall@3
  - cosine_recall@5
  - cosine_recall@10
  - cosine_ndcg@10
  - cosine_mrr@10
  - cosine_map@100
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:693000
  - loss:MatryoshkaLoss
  - loss:MultipleNegativesRankingLoss
widget:
  - source_sentence: >-
      Paracrystalline materials are defined as having short and medium range
      ordering in their lattice (similar to the liquid crystal phases) but
      lacking crystal-like long-range ordering at least in one direction.
    sentences:
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Paracrystalline
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Øystein Dahle
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Makis Belevonis
  - source_sentence: >-
      Hạ Trạch is a commune ( xã ) and village in Bố Trạch District , Quảng Bình
      Province , in Vietnam .   Category : Populated places in Quang Binh
      Province  Category : Communes of Quang Binh Province
    sentences:
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: The Taill of how this forsaid Tod maid his Confessioun to Freir
        Wolf Waitskaith
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Hạ Trạch
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Tadaxa
  - source_sentence: >-
      The Golden Mosque (سنهرى مسجد, Sunehri Masjid) is a mosque in Old Delhi.
      It is located outside the southwestern corner of Delhi Gate of the Red
      Fort, opposite the Netaji Subhash Park.
    sentences:
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Algorithm
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Golden Mosque (Red Fort)
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Parnaso Español
  - source_sentence: >-
      Unibank, S.A. is one of Haiti's two largest private commercial banks. The
      bank was founded in 1993 by a group of Haitian investors and is the main
      company of "Groupe Financier National (GFN)". It opened its first office
      in July 1993 in downtown Port-au-Prince and has 50 branches throughout the
      country as of the end of 2016.
    sentences:
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Sky TG24
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Ghomijeh
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Unibank (Haiti)
  - source_sentence: >-
      The Tchaikovsky Symphony Orchestra is a Russian classical music orchestra
      established in 1930. It was founded as the Moscow Radio Symphony
      Orchestra, and served as the official symphony for the Soviet All-Union
      Radio network. Following the dissolution of the, Soviet Union in 1991, the
      orchestra was renamed in 1993 by the Russian Ministry of Culture in
      recognition of the central role the music of Tchaikovsky plays in its
      repertoire. The current music director is Vladimir Fedoseyev, who has been
      in that position since 1974.
    sentences:
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Harald J.W. Mueller-Kirsten
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Sierra del Lacandón
      - >-
        Instruct: Given a web search query, retrieve relevant passages that
        answer the query.

        Query: Tchaikovsky Symphony Orchestra
model-index:
  - name: SentenceTransformer based on dunzhang/stella_en_1.5B_v5
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: cosine_accuracy@1
            value: 0.9457912457912457
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.9686868686868687
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.9750841750841751
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9818181818181818
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.9457912457912457
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.3228956228956229
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.195016835016835
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09818181818181818
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.9457912457912457
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.9686868686868687
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.9750841750841751
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9818181818181818
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.9641837379281919
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.9584885895997006
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.9590455638710143
            name: Cosine Map@100
          - type: cosine_accuracy@1
            value: 0.9447811447811448
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.9696969696969697
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.9754208754208754
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9824915824915825
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.9447811447811448
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.32323232323232326
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.19508417508417508
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09824915824915824
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.9447811447811448
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.9696969696969697
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.9754208754208754
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9824915824915825
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.9641053714591453
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.9581715301159749
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.9586773165340671
            name: Cosine Map@100
          - type: cosine_accuracy@1
            value: 0.9447811447811448
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.9673400673400674
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.9720538720538721
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9804713804713805
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.9447811447811448
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.32244668911335583
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.19441077441077437
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09804713804713805
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.9447811447811448
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.9673400673400674
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.9720538720538721
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9804713804713805
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.9628692157043424
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.9572219549997326
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.9577987764578036
            name: Cosine Map@100

SentenceTransformer based on dunzhang/stella_en_1.5B_v5

This is a sentence-transformers model finetuned from dunzhang/stella_en_1.5B_v5. 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: dunzhang/stella_en_1.5B_v5
  • Maximum Sequence Length: 8096 tokens
  • Output Dimensionality: 1024 tokens
  • Similarity Function: Cosine Similarity

Model Sources

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 8096, 'do_lower_case': False}) with Transformer model: Qwen2Model 
  (1): Pooling({'word_embedding_dimension': 1536, '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})
  (2): Dense({'in_features': 1536, 'out_features': 1024, 'bias': True, 'activation_function': 'torch.nn.modules.linear.Identity'})
)

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 = [
    'The Tchaikovsky Symphony Orchestra is a Russian classical music orchestra established in 1930. It was founded as the Moscow Radio Symphony Orchestra, and served as the official symphony for the Soviet All-Union Radio network. Following the dissolution of the, Soviet Union in 1991, the orchestra was renamed in 1993 by the Russian Ministry of Culture in recognition of the central role the music of Tchaikovsky plays in its repertoire. The current music director is Vladimir Fedoseyev, who has been in that position since 1974.',
    'Instruct: Given a web search query, retrieve relevant passages that answer the query.\nQuery: Tchaikovsky Symphony Orchestra',
    'Instruct: Given a web search query, retrieve relevant passages that answer the query.\nQuery: Sierra del Lacandón',
]
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

Information Retrieval

Metric Value
cosine_accuracy@1 0.9458
cosine_accuracy@3 0.9687
cosine_accuracy@5 0.9751
cosine_accuracy@10 0.9818
cosine_precision@1 0.9458
cosine_precision@3 0.3229
cosine_precision@5 0.195
cosine_precision@10 0.0982
cosine_recall@1 0.9458
cosine_recall@3 0.9687
cosine_recall@5 0.9751
cosine_recall@10 0.9818
cosine_ndcg@10 0.9642
cosine_mrr@10 0.9585
cosine_map@100 0.959

Information Retrieval

Metric Value
cosine_accuracy@1 0.9448
cosine_accuracy@3 0.9697
cosine_accuracy@5 0.9754
cosine_accuracy@10 0.9825
cosine_precision@1 0.9448
cosine_precision@3 0.3232
cosine_precision@5 0.1951
cosine_precision@10 0.0982
cosine_recall@1 0.9448
cosine_recall@3 0.9697
cosine_recall@5 0.9754
cosine_recall@10 0.9825
cosine_ndcg@10 0.9641
cosine_mrr@10 0.9582
cosine_map@100 0.9587

Information Retrieval

Metric Value
cosine_accuracy@1 0.9448
cosine_accuracy@3 0.9673
cosine_accuracy@5 0.9721
cosine_accuracy@10 0.9805
cosine_precision@1 0.9448
cosine_precision@3 0.3224
cosine_precision@5 0.1944
cosine_precision@10 0.098
cosine_recall@1 0.9448
cosine_recall@3 0.9673
cosine_recall@5 0.9721
cosine_recall@10 0.9805
cosine_ndcg@10 0.9629
cosine_mrr@10 0.9572
cosine_map@100 0.9578

Training Details

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: steps
  • per_device_eval_batch_size: 4
  • gradient_accumulation_steps: 4
  • learning_rate: 2e-05
  • max_steps: 1500
  • lr_scheduler_type: cosine
  • warmup_ratio: 0.1
  • warmup_steps: 5
  • bf16: True
  • tf32: True
  • optim: adamw_torch_fused
  • gradient_checkpointing: True
  • gradient_checkpointing_kwargs: {'use_reentrant': False}
  • batch_sampler: no_duplicates

All Hyperparameters

Click to expand
  • overwrite_output_dir: False
  • do_predict: False
  • eval_strategy: steps
  • prediction_loss_only: True
  • per_device_train_batch_size: 8
  • per_device_eval_batch_size: 4
  • per_gpu_train_batch_size: None
  • per_gpu_eval_batch_size: None
  • gradient_accumulation_steps: 4
  • eval_accumulation_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: 3.0
  • max_steps: 1500
  • lr_scheduler_type: cosine
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.1
  • warmup_steps: 5
  • 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: True
  • fp16: False
  • fp16_opt_level: O1
  • half_precision_backend: auto
  • bf16_full_eval: False
  • fp16_full_eval: False
  • tf32: True
  • local_rank: 0
  • ddp_backend: None
  • tpu_num_cores: None
  • tpu_metrics_debug: False
  • debug: []
  • dataloader_drop_last: True
  • 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_fused
  • 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: True
  • gradient_checkpointing_kwargs: {'use_reentrant': False}
  • 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
  • batch_sampler: no_duplicates
  • multi_dataset_batch_sampler: proportional

Training Logs

Epoch Step Training Loss loss cosine_map@100
0.0185 100 0.4835 0.0751 0.9138
0.0369 200 0.0646 0.0590 0.9384
0.0554 300 0.0594 0.0519 0.9462
0.0739 400 0.0471 0.0483 0.9514
0.0924 500 0.0524 0.0455 0.9531
0.1108 600 0.0435 0.0397 0.9546
0.1293 700 0.0336 0.0394 0.9549
0.1478 800 0.0344 0.0374 0.9565
0.1662 900 0.0393 0.0361 0.9568
0.1847 1000 0.0451 0.0361 0.9578

Framework Versions

  • Python: 3.10.12
  • Sentence Transformers: 3.0.1
  • Transformers: 4.41.2
  • PyTorch: 2.2.0+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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning}, 
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

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}
}