Hugging Face's logo

jvanhoof
/
all-miniLM-L6-v2-mE5s-distilled-ja-en-64K-vocab-mean-17M-test

Sentence Similarity
sentence-transformers
Safetensors
bert
feature-extraction
Generated from Trainer
dataset_size:16681525
loss:MSELoss
Eval Results
text-embeddings-inference
Model card Files Community
all-miniLM-L6-v2-mE5s-distilled-ja-en-64K-vocab-mean-17M-test / README.md
jvanhoof's picture
jvanhoof
Upload folder using huggingface_hub
98686d9 verified
preview code
|
raw
history blame contribute delete
43.2 kB
metadata
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:16681525
  - loss:MSELoss
base_model: sentence-transformers/all-MiniLM-L6-v2
widget:
  - source_sentence: 舞台デザイン Scott Pask
    sentences:
      - Shore is a fine strappin’ man for shore!”
      - ３年間、空軍に所属。
      - >-
        近年、カイン、カチンおよびシャン州などいくつかの少数民族居住地域では、紛争が再発、あるいは激化しており、ビルマ国軍による市民への広範で組織的な人権侵害が続いていると報告されています。
  - source_sentence: で再放送があると　またまたまたまたまた見てしまう・・・１９９１年の今頃（冬）の作品。
    sentences:
      - これを書いたコラムニストは言う。
      - >-
        How might one who is battling negative feelings gain a more positive
        viewpoint?
      - 彼女は彼に何もしてあげるつもりはない。
  - source_sentence: When you feel something in your heart, it is a true feeling.
    sentences:
      - できたコースターなど。
      - >-
        But reflecting that it would be a good two hours at least before his
        arrival she concluded to go up the road to Robert Bell’s and tell them
        the news.
      - ‘ THERE they are again!
  - source_sentence: あなた ​ は ​ イエス ​ の ​ 名 ​ に ​ よっ ​ て『ひざ ​ を ​ かがめる』でしょ ​ う ​ か。
    sentences:
      - >-
        主治 ​ 医 ​ は ， やっ ​ て ​ みる ​ 価値 ​ は ​ ある ​ と ​ 判断 ​ し ， それら ​ の ​ 抗生物質 ​ を
        ​ 処方 ​ し ​ て ​ くれ ​ まし ​ た。
      - So he must have been due a rest in in 2015, surely?
      - わたしたちをあなたの愛の道具としてくださいますように。
  - source_sentence: No.
    sentences:
      - >-
        Today, while mankind as a whole can make no headway in abolishing war,
        there is again a group of people who have achieved this same remarkable
        goal.
      - >-
        They are not among those who are happy because of being conscious of
        their spiritual need. ’
      - >-
        Do not Jesus ’ words help us to appreciate what Jehovah really wants
        from us?
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - pearson_cosine
  - spearman_cosine
model-index:
  - name: SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: STSbenchmark en test
          type: STSbenchmark-en-test
        metrics:
          - type: pearson_cosine
            value: 0.7900115146348707
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.8206080071379148
            name: Spearman Cosine
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: JSTS validation
          type: JSTS-validation
        metrics:
          - type: pearson_cosine
            value: 0.830334965166667
            name: Pearson Cosine
          - type: spearman_cosine
            value: 0.7832534974732547
            name: Spearman Cosine

SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. 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: sentence-transformers/all-MiniLM-L6-v2
  • Maximum Sequence Length: 128 tokens
  • Output Dimensionality: 384 dimensions
  • Similarity Function: Cosine Similarity

Model Sources

Full Model Architecture 

SentenceTransformer(
  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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): 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 = [
    'No.',
    'Today, while mankind as a whole can make no headway in abolishing war, there is again a group of people who have achieved this same remarkable goal.',
    'Do not Jesus ’ words help us to appreciate what Jehovah really wants from us?',
]
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

Semantic Similarity

Metric STSbenchmark-en-test JSTS-validation
pearson_cosine 0.79 0.8303
spearman_cosine 0.8206 0.7833

Training Details

Training Dataset

Unnamed Dataset

  • Size: 16,681,525 training samples
  • Columns: english and label
  • Approximate statistics based on the first 1000 samples:
    english label
    type string list
    details
    • min: 4 tokens
    • mean: 24.06 tokens
    • max: 128 tokens
    • size: 384 elements
  • Samples:
    english label
    Before being used as a Kingdom Hall, it is cleared of any relics of false worship. [0.07804396003484726, -0.03384765610098839, -0.02494540810585022, -0.07025666534900665, 0.07451561838388443, ...]
    That girl always appeared from somewhere and followed me around. [0.032411519438028336, -0.014580070041120052, -0.005011037457734346, -0.030591191723942757, 0.10543882101774216, ...]
    いいえ そういう高い創造力を 保ち続ける科学者は 平均すると 最初の研究論文100編で 43回もトピックを変えていました [0.041001513600349426, -0.0013687843456864357, -0.017593644559383392, -0.03798963502049446, 0.07939894497394562, ...]
  • Loss: MSELoss

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: steps
  • per_device_train_batch_size: 512
  • per_device_eval_batch_size: 512
  • gradient_accumulation_steps: 2
  • learning_rate: 0.0003
  • num_train_epochs: 8
  • warmup_ratio: 0.15
  • bf16: True
  • dataloader_num_workers: 8
  • 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: 512
  • per_device_eval_batch_size: 512
  • per_gpu_train_batch_size: None
  • per_gpu_eval_batch_size: None
  • gradient_accumulation_steps: 2
  • eval_accumulation_steps: None
  • torch_empty_cache_steps: None
  • learning_rate: 0.0003
  • weight_decay: 0.0
  • adam_beta1: 0.9
  • adam_beta2: 0.999
  • adam_epsilon: 1e-08
  • max_grad_norm: 1.0
  • num_train_epochs: 8
  • max_steps: -1
  • lr_scheduler_type: linear
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.15
  • 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: True
  • 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: 8
  • 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}
  • tp_size: 0
  • 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: None
  • hub_always_push: False
  • 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: False
  • neftune_noise_alpha: None
  • optim_target_modules: None
  • batch_eval_metrics: False
  • eval_on_start: False
  • use_liger_kernel: False
  • eval_use_gather_object: False
  • average_tokens_across_devices: False
  • prompts: None
  • batch_sampler: batch_sampler
  • multi_dataset_batch_sampler: proportional

Training Logs

Click to expand
Epoch Step Training Loss STSbenchmark-en-test_spearman_cosine JSTS-validation_spearman_cosine
0.0307 500 0.004 - -
0.0614 1000 0.0013 - -
0.0921 1500 0.0011 - -
0.1228 2000 0.0011 0.1715 0.2170
0.1535 2500 0.0011 - -
0.1842 3000 0.001 - -
0.2148 3500 0.001 - -
0.2455 4000 0.001 0.3243 0.3289
0.2762 4500 0.0009 - -
0.3069 5000 0.0009 - -
0.3376 5500 0.0008 - -
0.3683 6000 0.0008 0.4725 0.5123
0.3990 6500 0.0007 - -
0.4297 7000 0.0007 - -
0.4604 7500 0.0007 - -
0.4911 8000 0.0006 0.5669 0.6228
0.5218 8500 0.0006 - -
0.5525 9000 0.0006 - -
0.5831 9500 0.0006 - -
0.6138 10000 0.0005 0.6407 0.6840
0.6445 10500 0.0005 - -
0.6752 11000 0.0005 - -
0.7059 11500 0.0005 - -
0.7366 12000 0.0005 0.6953 0.7212
0.7673 12500 0.0005 - -
0.7980 13000 0.0004 - -
0.8287 13500 0.0004 - -
0.8594 14000 0.0004 0.7358 0.7510
0.8901 14500 0.0004 - -
0.9208 15000 0.0004 - -
0.9514 15500 0.0004 - -
0.9821 16000 0.0004 0.7550 0.7612
1.0128 16500 0.0003 - -
1.0435 17000 0.0003 - -
1.0742 17500 0.0003 - -
1.1049 18000 0.0003 0.7807 0.7739
1.1356 18500 0.0003 - -
1.1663 19000 0.0003 - -
1.1970 19500 0.0003 - -
1.2277 20000 0.0003 0.8003 0.7781
1.2584 20500 0.0003 - -
1.2891 21000 0.0003 - -
1.3197 21500 0.0003 - -
1.3504 22000 0.0002 0.8050 0.7780
1.3811 22500 0.0002 - -
1.4118 23000 0.0002 - -
1.4425 23500 0.0002 - -
1.4732 24000 0.0002 0.8098 0.7788
1.5039 24500 0.0002 - -
1.5346 25000 0.0002 - -
1.5653 25500 0.0002 - -
1.5960 26000 0.0002 0.8147 0.7804
1.6267 26500 0.0002 - -
1.6574 27000 0.0002 - -
1.6880 27500 0.0002 - -
1.7187 28000 0.0002 0.8136 0.7762
1.7494 28500 0.0002 - -
1.7801 29000 0.0002 - -
1.8108 29500 0.0002 - -
1.8415 30000 0.0002 0.8138 0.7784
1.8722 30500 0.0002 - -
1.9029 31000 0.0002 - -
1.9336 31500 0.0002 - -
1.9643 32000 0.0002 0.8161 0.7812
1.9950 32500 0.0002 - -
2.0257 33000 0.0002 - -
2.0564 33500 0.0002 - -
2.0870 34000 0.0002 0.8140 0.7790
2.1177 34500 0.0002 - -
2.1484 35000 0.0002 - -
2.1791 35500 0.0002 - -
2.2098 36000 0.0002 0.8136 0.7815
2.2405 36500 0.0002 - -
2.2712 37000 0.0002 - -
2.3019 37500 0.0002 - -
2.3326 38000 0.0002 0.8165 0.7811
2.3633 38500 0.0002 - -
2.3940 39000 0.0002 - -
2.4247 39500 0.0002 - -
2.4553 40000 0.0002 0.8157 0.7822
2.4860 40500 0.0002 - -
2.5167 41000 0.0002 - -
2.5474 41500 0.0002 - -
2.5781 42000 0.0002 0.8149 0.7819
2.6088 42500 0.0002 - -
2.6395 43000 0.0002 - -
2.6702 43500 0.0002 - -
2.7009 44000 0.0002 0.8167 0.7812
2.7316 44500 0.0002 - -
2.7623 45000 0.0002 - -
2.7930 45500 0.0002 - -
2.8236 46000 0.0002 0.8146 0.7829
2.8543 46500 0.0002 - -
2.8850 47000 0.0002 - -
2.9157 47500 0.0002 - -
2.9464 48000 0.0002 0.8162 0.7817
2.9771 48500 0.0002 - -
3.0078 49000 0.0002 - -
3.0385 49500 0.0002 - -
3.0692 50000 0.0002 0.8155 0.7818
3.0999 50500 0.0002 - -
3.1306 51000 0.0002 - -
3.1613 51500 0.0002 - -
3.1919 52000 0.0002 0.8160 0.7813
3.2226 52500 0.0001 - -
3.2533 53000 0.0001 - -
3.2840 53500 0.0001 - -
3.3147 54000 0.0001 0.8190 0.7822
3.3454 54500 0.0001 - -
3.3761 55000 0.0001 - -
3.4068 55500 0.0001 - -
3.4375 56000 0.0001 0.8172 0.7834
3.4682 56500 0.0001 - -
3.4989 57000 0.0001 - -
3.5296 57500 0.0001 - -
3.5602 58000 0.0001 0.8175 0.7822
3.5909 58500 0.0001 - -
3.6216 59000 0.0001 - -
3.6523 59500 0.0001 - -
3.6830 60000 0.0001 0.8188 0.7827
3.7137 60500 0.0001 - -
3.7444 61000 0.0001 - -
3.7751 61500 0.0001 - -
3.8058 62000 0.0001 0.8162 0.7829
3.8365 62500 0.0001 - -
3.8672 63000 0.0001 - -
3.8979 63500 0.0001 - -
3.9285 64000 0.0001 0.8185 0.7814
3.9592 64500 0.0001 - -
3.9899 65000 0.0001 - -
4.0206 65500 0.0001 - -
4.0513 66000 0.0001 0.8174 0.7839
4.0820 66500 0.0001 - -
4.1127 67000 0.0001 - -
4.1434 67500 0.0001 - -
4.1741 68000 0.0001 0.8174 0.7827
4.2048 68500 0.0001 - -
4.2355 69000 0.0001 - -
4.2662 69500 0.0001 - -
4.2969 70000 0.0001 0.8166 0.7833
4.3275 70500 0.0001 - -
4.3582 71000 0.0001 - -
4.3889 71500 0.0001 - -
4.4196 72000 0.0001 0.8190 0.7837
4.4503 72500 0.0001 - -
4.4810 73000 0.0001 - -
4.5117 73500 0.0001 - -
4.5424 74000 0.0001 0.8188 0.7837
4.5731 74500 0.0001 - -
4.6038 75000 0.0001 - -
4.6345 75500 0.0001 - -
4.6652 76000 0.0001 0.8186 0.7829
4.6958 76500 0.0001 - -
4.7265 77000 0.0001 - -
4.7572 77500 0.0001 - -
4.7879 78000 0.0001 0.8188 0.7839
4.8186 78500 0.0001 - -
4.8493 79000 0.0001 - -
4.8800 79500 0.0001 - -
4.9107 80000 0.0001 0.8187 0.7830
4.9414 80500 0.0001 - -
4.9721 81000 0.0001 - -
5.0028 81500 0.0001 - -
5.0335 82000 0.0001 0.8200 0.7829
5.0641 82500 0.0001 - -
5.0948 83000 0.0001 - -
5.1255 83500 0.0001 - -
5.1562 84000 0.0001 0.8195 0.7842
5.1869 84500 0.0001 - -
5.2176 85000 0.0001 - -
5.2483 85500 0.0001 - -
5.2790 86000 0.0001 0.8205 0.7835
5.3097 86500 0.0001 - -
5.3404 87000 0.0001 - -
5.3711 87500 0.0001 - -
5.4018 88000 0.0001 0.8201 0.7840
5.4324 88500 0.0001 - -
5.4631 89000 0.0001 - -
5.4938 89500 0.0001 - -
5.5245 90000 0.0001 0.8199 0.7830
5.5552 90500 0.0001 - -
5.5859 91000 0.0001 - -
5.6166 91500 0.0001 - -
5.6473 92000 0.0001 0.8204 0.7830
5.6780 92500 0.0001 - -
5.7087 93000 0.0001 - -
5.7394 93500 0.0001 - -
5.7701 94000 0.0001 0.8206 0.7832
5.8007 94500 0.0001 - -
5.8314 95000 0.0001 - -
5.8621 95500 0.0001 - -
5.8928 96000 0.0001 0.8212 0.7833
5.9235 96500 0.0001 - -
5.9542 97000 0.0001 - -
5.9849 97500 0.0001 - -
6.0156 98000 0.0001 0.8210 0.7839
6.0463 98500 0.0001 - -
6.0770 99000 0.0001 - -
6.1077 99500 0.0001 - -
6.1384 100000 0.0001 0.8206 0.7850
6.1691 100500 0.0001 - -
6.1997 101000 0.0001 - -
6.2304 101500 0.0001 - -
6.2611 102000 0.0001 0.8200 0.7839
6.2918 102500 0.0001 - -
6.3225 103000 0.0001 - -
6.3532 103500 0.0001 - -
6.3839 104000 0.0001 0.8197 0.7827
6.4146 104500 0.0001 - -
6.4453 105000 0.0001 - -
6.4760 105500 0.0001 - -
6.5067 106000 0.0001 0.8197 0.7837
6.5374 106500 0.0001 - -
6.5680 107000 0.0001 - -
6.5987 107500 0.0001 - -
6.6294 108000 0.0001 0.8214 0.7837
6.6601 108500 0.0001 - -
6.6908 109000 0.0001 - -
6.7215 109500 0.0001 - -
6.7522 110000 0.0001 0.8195 0.7823
6.7829 110500 0.0001 - -
6.8136 111000 0.0001 - -
6.8443 111500 0.0001 - -
6.8750 112000 0.0001 0.8203 0.7836
6.9057 112500 0.0001 - -
6.9363 113000 0.0001 - -
6.9670 113500 0.0001 - -
6.9977 114000 0.0001 0.8213 0.7832
7.0284 114500 0.0001 - -
7.0591 115000 0.0001 - -
7.0898 115500 0.0001 - -
7.1205 116000 0.0001 0.8213 0.7844
7.1512 116500 0.0001 - -
7.1819 117000 0.0001 - -
7.2126 117500 0.0001 - -
7.2433 118000 0.0001 0.8202 0.7831
7.2740 118500 0.0001 - -
7.3046 119000 0.0001 - -
7.3353 119500 0.0001 - -
7.3660 120000 0.0001 0.8203 0.7837
7.3967 120500 0.0001 - -
7.4274 121000 0.0001 - -
7.4581 121500 0.0001 - -
7.4888 122000 0.0001 0.8207 0.7833
7.5195 122500 0.0001 - -
7.5502 123000 0.0001 - -
7.5809 123500 0.0001 - -
7.6116 124000 0.0001 0.8202 0.7835
7.6423 124500 0.0001 - -
7.6729 125000 0.0001 - -
7.7036 125500 0.0001 - -
7.7343 126000 0.0001 0.8209 0.7832
7.7650 126500 0.0001 - -
7.7957 127000 0.0001 - -
7.8264 127500 0.0001 - -
7.8571 128000 0.0001 0.8206 0.7837
7.8878 128500 0.0001 - -
7.9185 129000 0.0001 - -
7.9492 129500 0.0001 - -
7.9799 130000 0.0001 0.8206 0.7833

Framework Versions

  • Python: 3.10.16
  • Sentence Transformers: 3.3.1
  • Transformers: 4.51.3
  • PyTorch: 2.5.1+cu124
  • Accelerate: 1.2.1
  • Datasets: 3.2.0
  • Tokenizers: 0.21.1

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",
}

MSELoss 

@inproceedings{reimers-2020-multilingual-sentence-bert,
    title = "Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2020",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/2004.09813",
}