Hugging Face's logo

lucian-li
/
my_new_model

Sentence Similarity
sentence-transformers
Safetensors
new
feature-extraction
Generated from Trainer
dataset_size:583058
loss:MultipleNegativesRankingLoss
custom_code
text-embeddings-inference
Model card Files Community
my_new_model / README.md
lucian-li's picture
lucian-li
Add new SentenceTransformer model
feae91a verified
preview code
|
raw
history blame contribute delete
38.5 kB
metadata
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:583058
  - loss:MultipleNegativesRankingLoss
base_model: Alibaba-NLP/gte-multilingual-base
widget:
  - source_sentence: |-
      Pre-Emphasis (PE)
      A pre-emphasis filter is applied to the framed offset-free input signal:

      )1
      (
    sentences:
      - >-
        Windowing (W)

        A Hamming window of length N is applied to the output of the
        pre-emphasis block:


        (

        )

        N

        n

        n

        s

        N

        n

        n

        s

        pe

        w

        ≤

        ≤

        ×

        

        

        

        

        

        

        

        

        

        

        

        

        −

        −

        ×

        −

        =
      - >-
        Group or broadcast call, called mobile stations (GSM only)

        Within each set of voice group call or voice broadcast call attributes
        stored in the GCR as defined in 3GPP TS 43.068

        and 3GPP TS 43.069, respectively, a priority level is included if eMLPP
        is applied. The priority level will be provided

        by the GCR to the MSC together with the call attributes.

        The priority level shall be indicated together with the related
        notification messages and treated in the mobile station as

        defined in 3GPP TS 43.0
      - >-
        Description of the access technology indicator mechanism

        This clause describes the mechanisms that can be employed to indicate
        access technology specific dependencies in a

        multi-access technology environment.

        There are cases where toolkit applications need to know which access
        technology the terminal is currently in so that it

        can issue access technology dependent commands as well as determine that
        the response to a particular command is

        technology dependent. Setting up the event, ACCESS TECHNOL
  - source_sentence: >-
      Distribution of DL delay between NG-RAN and UE

      a) This measurement provides the distribution of DL packet delay between
      NG-RAN and UE, which is the delay

      incurred in NG-RAN (including the delay at gNB-CU-UP, on F1-U and on
      gNB-DU) and the delay over Uu

      interface. This measurement is split into subcounters per 5QI and
      subcounters per S-NSSAI.

      b) DER (n=1).


      ETSI

      ETSI TS 128 552 V16.18.0 (2024-08)
    sentences:
      - >-
        Distribution of UL delay between NG-RAN and UE

        a) This measurement provides the distribution of UL packet delay between
        NG-RAN and UE, which is the delay

        incurred in NG-RAN (including the delay at gNB-CU-UP, on F1-U and on
        gNB-DU) and the delay over Uu

        interface. This measurement is split into subcounters per 5QI and
        subcounters per S-NSSAI.

        b) DER (n=1).

        c) The measurement is obtained by the following method:


        The gNB performs the GTP PDU packet delay measurement for QoS monitoring
        per the GTP 
      - >-
        Subscriber data

        Subscription to MExE services shall be logically separate to
        subscription of network services. A subscriber may have a

        MExE subscription to multiple MExE service providers. It may also be
        possible for the subscriber to interrogate such

        subscription registration (with a suitable means of authorisation),
        depending on PLMN support.
      - >-
        MSC for LMU Control

        When a control message has to be routed to an LMU from an SMLC, the SMLC
        addresses the serving MSC for the

        LMU using an E.164 address.


        ETSI

        ETSI TS 129 002 V10.6.0 (2012-04)
  - source_sentence: >-
      Enter SMS Block Mode Protocol +CESP

      Table 3.2.4-1: +CESP Action Command Syntax

      Command

      Possible response(s)

      +CESP


      +CESP=?



      Description

      Execution command sets the TA in SMS block protocol mode. The TA shall
      return OK (or 0) to confirm acceptance of

      the command prior to entering the block mode (see clause 2.1.1). The final
      result code OK (or 0) shall be returned when

      the block mode is exited.

      NOTE:

      Commands following +CESP in the AT command line must not be processed by
      the TA.

      Implementation

      Ma
    sentences:
      - |-
        SGSN
        To support NBIFOM, the SGSN needs to be capable to:

        ETSI
        ETSI TS 123 161 V14.0.0 (2017-05)
      - >-
        Message Service Failure Result Code +CMS ERROR

        Final result code +CMS ERROR: <err> indicates an error related to mobile
        equipment or network. The operation is

        similar to ERROR final result code. None of the following commands in
        the same command line is executed. Neither

        ERROR nor OK final result code shall be returned. ERROR is returned
        normally when error is related to syntax or invalid

        parameters.

        Defined Values

        <err> values used by common messaging commands:
      - |-
        C
        C
        -
        -
        P
        Service Priority Level
  - source_sentence: >-
      Definition

      Cell synchronization accuracy is defined as the maximum deviation in frame
      start times between any pair of cells on the

      same frequency that have overlapping coverage areas.
    sentences:
      - |-
        Minimum requirements
        The cell synchronization accuracy shall be better than or equal to 3μs.
      - "Subsequent Inter-MSC Handover to third MSC\nWhen a Mobile Station is being handed over to a third MSC, the procedure (described in GSM 03.09)\ndoes require one specific interworking case in MSC-A (figure 20) between E-Interface from MSC-B and E-\nInterface from MSC-B' other than the combination of the ones described in the chapter 4.5.1 and 4.5.2.\n%66\x10$\x03\x03\x03\x03\x03\x0306&\x10%\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x0306&\x10$\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x0306&\x10%\n_\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03_\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03_\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03_\n_+$1'29(5\x03\x03\x03\x03_\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03\x03_\x03\x03\x03\x03"
      - >-
        DL Total PRB Usage

        a) This measurement provides the total usage (in percentage) of physical
        resource blocks (PRBs) on the downlink

        for any purpose.

        b) SI

        c) This measurement is obtained as:

        

        

        

        

        ∗

        =
  - source_sentence: Carrier aggregation measurement accuracy
    sentences:
      - >-
        PUCCH / PUSCH / SRS time mask

        The PUCCH/PUSCH/SRS time mask defines the observation period between
        sounding reference symbol (SRS) and an

        adjacent PUSCH/PUCCH symbol and subsequent sub-frame.

        There are no additional requirements on UE transmit power beyond that
        which is required in subclause 6.2.2 and

        subclause 6.6.2.3


        ETSI

        ETSI TS 136 101 V9.16.0 (2013-07)
      - >-
        Reference Signal Time Difference (RSTD) Measurement Accuracy

        Requirements for Carrier Aggregation

        A.8

        UE Measurements Procedures

        A.9

        Measurement Performance Requirements

        NOTE:

        Only requirements and test cases in this table defined for inter-band
        carrier aggregation shall apply.



        ETSI

        ETSI TS 136 307 V10.17.0 (2016-01)
      - >-
        Operator control

        Three general architectures are candidates to offer energy savings
        functionalities:

        Distributed, NM-Centralized, EM-Centralized as defined in TS 32.500 [6].

        Energy savings in cells can be initiated in several different ways. Some
        of the mechanisms are:

        For NM-centralized architecture

        -

        IRPManager instructs the cells to move to energySaving state (e.g.
        according to a schedule determined by

        network statistics) , configures trigger points (e.g. load threshold
        crossing) when it want
pipeline_tag: sentence-similarity
library_name: sentence-transformers

SentenceTransformer based on Alibaba-NLP/gte-multilingual-base

This is a sentence-transformers model finetuned from Alibaba-NLP/gte-multilingual-base. 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: Alibaba-NLP/gte-multilingual-base
  • Maximum Sequence Length: 8192 tokens
  • Output Dimensionality: 768 dimensions
  • Similarity Function: Cosine Similarity

Model Sources

Full Model Architecture 

SentenceTransformer(
  (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: NewModel 
  (1): Pooling({'word_embedding_dimension': 768, '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("lucian-li/my_new_model")
# Run inference
sentences = [
    'Carrier aggregation measurement accuracy',
    'Reference Signal Time Difference (RSTD) Measurement Accuracy\nRequirements for Carrier Aggregation\nA.8\nUE Measurements Procedures\nA.9\nMeasurement Performance Requirements\nNOTE:\nOnly requirements and test cases in this table defined for inter-band carrier aggregation shall apply.\n\n\nETSI\nETSI TS 136 307 V10.17.0 (2016-01)',
    'Operator control\nThree general architectures are candidates to offer energy savings functionalities:\nDistributed, NM-Centralized, EM-Centralized as defined in TS 32.500 [6].\nEnergy savings in cells can be initiated in several different ways. Some of the mechanisms are:\nFor NM-centralized architecture\n-\nIRPManager instructs the cells to move to energySaving state (e.g. according to a schedule determined by\nnetwork statistics) , configures trigger points (e.g. load threshold crossing) when it want',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

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

Training Details

Training Dataset

Unnamed Dataset

  • Size: 583,058 training samples
  • Columns: anchor and positive
  • Approximate statistics based on the first 1000 samples:
    anchor positive
    type string string
    details
    • min: 7 tokens
    • mean: 85.73 tokens
    • max: 229 tokens
    • min: 7 tokens
    • mean: 85.86 tokens
    • max: 229 tokens
  • Samples:
    anchor positive
    Triggering Optimization Function (TG_F)
    This functional bloc supports the following functions: [SO2], [SO3].    		 Optimization Fallback Function (O_FB_F)
    This functional bloc supports the following functions: [SO7], [SO9], [SO10].
    Optimization Fallback Function (O_FB_F)
    This functional bloc supports the following functions: [SO7], [SO9], [SO10].    		 Self-Optimization Progress Update Function (SO_PGS_UF)
    This function updates the self-optimization progress and important events to the operator: [SO11]
    Self-Optimization Progress Update Function (SO_PGS_UF)
    This function updates the self-optimization progress and important events to the operator: [SO11]    		 NRM IRP Update Function (NRM_UF)
    This function updates the E-UTRAN and EPC NRM IRP with the optimization modification if needed.
  • Loss: MultipleNegativesRankingLoss with these parameters:
    {
    "scale": 20.0,
    "similarity_fct": "cos_sim"
}

Training Hyperparameters

Non-Default Hyperparameters

  • per_device_train_batch_size: 11
  • num_train_epochs: 1
  • warmup_ratio: 0.1

All Hyperparameters

Click to expand
  • overwrite_output_dir: False
  • do_predict: False
  • eval_strategy: no
  • prediction_loss_only: True
  • per_device_train_batch_size: 11
  • per_device_eval_batch_size: 8
  • 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: 1.0
  • num_train_epochs: 1
  • max_steps: -1
  • lr_scheduler_type: linear
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.1
  • 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}
  • 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
0.0019 100 0.8198
0.0038 200 0.7651
0.0057 300 0.6659
0.0075 400 0.6404
0.0094 500 0.5638
0.0113 600 0.5184
0.0132 700 0.448
0.0151 800 0.4464
0.0170 900 0.3461
0.0189 1000 0.3731
0.0208 1100 0.343
0.0226 1200 0.3557
0.0245 1300 0.3623
0.0264 1400 0.2941
0.0283 1500 0.3153
0.0302 1600 0.2724
0.0321 1700 0.2702
0.0340 1800 0.2934
0.0358 1900 0.2255
0.0377 2000 0.2519
0.0396 2100 0.2424
0.0415 2200 0.1883
0.0434 2300 0.2428
0.0453 2400 0.2212
0.0472 2500 0.1862
0.0491 2600 0.2451
0.0509 2700 0.2336
0.0528 2800 0.225
0.0547 2900 0.2154
0.0566 3000 0.1907
0.0585 3100 0.2514
0.0604 3200 0.2082
0.0623 3300 0.2076
0.0641 3400 0.1818
0.0660 3500 0.1688
0.0679 3600 0.2261
0.0698 3700 0.2108
0.0717 3800 0.1732
0.0736 3900 0.1764
0.0755 4000 0.1481
0.0773 4100 0.1687
0.0792 4200 0.1897
0.0811 4300 0.1685
0.0830 4400 0.1915
0.0849 4500 0.2013
0.0868 4600 0.1701
0.0887 4700 0.2006
0.0906 4800 0.2006
0.0924 4900 0.1617
0.0943 5000 0.1406
0.0962 5100 0.1456
0.0981 5200 0.1703
0.1000 5300 0.1464
0.1019 5400 0.1803
0.1038 5500 0.1346
0.1056 5600 0.134
0.1075 5700 0.1567
0.1094 5800 0.163
0.1113 5900 0.1544
0.1132 6000 0.1648
0.1151 6100 0.1505
0.1170 6200 0.1231
0.1189 6300 0.1591
0.1207 6400 0.1533
0.1226 6500 0.1376
0.1245 6600 0.1473
0.1264 6700 0.1405
0.1283 6800 0.141
0.1302 6900 0.1105
0.1321 7000 0.1712
0.1339 7100 0.1534
0.1358 7200 0.1578
0.1377 7300 0.1101
0.1396 7400 0.128
0.1415 7500 0.1679
0.1434 7600 0.1592
0.1453 7700 0.1383
0.1472 7800 0.1274
0.1490 7900 0.1616
0.1509 8000 0.1617
0.1528 8100 0.1361
0.1547 8200 0.1268
0.1566 8300 0.1286
0.1585 8400 0.1253
0.1604 8500 0.1157
0.1622 8600 0.1499
0.1641 8700 0.1398
0.1660 8800 0.1188
0.1679 8900 0.1103
0.1698 9000 0.1217
0.1717 9100 0.1144
0.1736 9200 0.1203
0.1755 9300 0.1074
0.1773 9400 0.1145
0.1792 9500 0.1035
0.1811 9600 0.1406
0.1830 9700 0.1465
0.1849 9800 0.1169
0.1868 9900 0.1115
0.1887 10000 0.1207
0.1905 10100 0.1191
0.1924 10200 0.1099
0.1943 10300 0.1309
0.1962 10400 0.1092
0.1981 10500 0.1075
0.2000 10600 0.1174
0.2019 10700 0.1103
0.2038 10800 0.1077
0.2056 10900 0.0844
0.2075 11000 0.1093
0.2094 11100 0.1428
0.2113 11200 0.0928
0.2132 11300 0.1039
0.2151 11400 0.1436
0.2170 11500 0.1197
0.2188 11600 0.1249
0.2207 11700 0.0856
0.2226 11800 0.1126
0.2245 11900 0.1028
0.2264 12000 0.0988
0.2283 12100 0.1031
0.2302 12200 0.101
0.2320 12300 0.1188
0.2339 12400 0.0908
0.2358 12500 0.069
0.2377 12600 0.1099
0.2396 12700 0.1227
0.2415 12800 0.0794
0.2434 12900 0.0969
0.2453 13000 0.0864
0.2471 13100 0.1193
0.2490 13200 0.0824
0.2509 13300 0.12
0.2528 13400 0.0928
0.2547 13500 0.1126
0.2566 13600 0.0912
0.2585 13700 0.1126
0.2603 13800 0.078
0.2622 13900 0.0715
0.2641 14000 0.1095
0.2660 14100 0.089
0.2679 14200 0.0926
0.2698 14300 0.086
0.2717 14400 0.1115
0.2736 14500 0.0996
0.2754 14600 0.1014
0.2773 14700 0.1033
0.2792 14800 0.0732
0.2811 14900 0.0994
0.2830 15000 0.0872
0.2849 15100 0.0923
0.2868 15200 0.111
0.2886 15300 0.0891
0.2905 15400 0.0868
0.2924 15500 0.0773
0.2943 15600 0.0918
0.2962 15700 0.0726
0.2981 15800 0.0951
0.3000 15900 0.0835
0.3019 16000 0.083
0.3037 16100 0.095
0.3056 16200 0.0722
0.3075 16300 0.1061
0.3094 16400 0.0902
0.3113 16500 0.0978
0.3132 16600 0.0983
0.3151 16700 0.0808
0.3169 16800 0.0758
0.3188 16900 0.071
0.3207 17000 0.0918
0.3226 17100 0.1011
0.3245 17200 0.079
0.3264 17300 0.0992
0.3283 17400 0.1089
0.3302 17500 0.0904
0.3320 17600 0.0956
0.3339 17700 0.0747
0.3358 17800 0.0961
0.3377 17900 0.0923
0.3396 18000 0.1114
0.3415 18100 0.0689
0.3434 18200 0.1308
0.3452 18300 0.0923
0.3471 18400 0.0756
0.3490 18500 0.0842
0.3509 18600 0.0859
0.3528 18700 0.0903
0.3547 18800 0.084
0.3566 18900 0.0923
0.3584 19000 0.0848
0.3603 19100 0.0812
0.3622 19200 0.0872
0.3641 19300 0.083
0.3660 19400 0.0826
0.3679 19500 0.101
0.3698 19600 0.0804
0.3717 19700 0.0676
0.3735 19800 0.0836
0.3754 19900 0.0711
0.3773 20000 0.0825
0.3792 20100 0.0835
0.3811 20200 0.0816
0.3830 20300 0.0812
0.3849 20400 0.0689
0.3867 20500 0.0627
0.3886 20600 0.0965
0.3905 20700 0.0632
0.3924 20800 0.0945
0.3943 20900 0.0923
0.3962 21000 0.0833
0.3981 21100 0.0537
0.4000 21200 0.0822
0.4018 21300 0.0684
0.4037 21400 0.0807
0.4056 21500 0.0945
0.4075 21600 0.0981
0.4094 21700 0.0748
0.4113 21800 0.0943
0.4132 21900 0.0709
0.4150 22000 0.0551
0.4169 22100 0.0679
0.4188 22200 0.0666
0.4207 22300 0.0976
0.4226 22400 0.0666
0.4245 22500 0.0651
0.4264 22600 0.0803
0.4283 22700 0.068
0.4301 22800 0.0541
0.4320 22900 0.0487
0.4339 23000 0.091
0.4358 23100 0.074
0.4377 23200 0.0733
0.4396 23300 0.0845
0.4415 23400 0.0823
0.4433 23500 0.0561
0.4452 23600 0.0508
0.4471 23700 0.074
0.4490 23800 0.0683
0.4509 23900 0.0797
0.4528 24000 0.0561
0.4547 24100 0.0744
0.4566 24200 0.0638
0.4584 24300 0.0633
0.4603 24400 0.062
0.4622 24500 0.0887
0.4641 24600 0.0908
0.4660 24700 0.0654
0.4679 24800 0.0522
0.4698 24900 0.0851
0.4716 25000 0.0763
0.4735 25100 0.0623
0.4754 25200 0.0712
0.4773 25300 0.0866
0.4792 25400 0.0812
0.4811 25500 0.0706
0.4830 25600 0.0734
0.4849 25700 0.068
0.4867 25800 0.111
0.4886 25900 0.0627
0.4905 26000 0.0459
0.4924 26100 0.0794
0.4943 26200 0.0547
0.4962 26300 0.0779
0.4981 26400 0.0609
0.4999 26500 0.0785
0.5018 26600 0.0722
0.5037 26700 0.0585
0.5056 26800 0.0572
0.5075 26900 0.0636
0.5094 27000 0.0642
0.5113 27100 0.0606
0.5131 27200 0.0725
0.5150 27300 0.0664
0.5169 27400 0.0933
0.5188 27500 0.0486
0.5207 27600 0.0514
0.5226 27700 0.0779
0.5245 27800 0.0614
0.5264 27900 0.0646
0.5282 28000 0.0606
0.5301 28100 0.0453
0.5320 28200 0.0749
0.5339 28300 0.0695
0.5358 28400 0.0897
0.5377 28500 0.0612
0.5396 28600 0.0542
0.5414 28700 0.0504
0.5433 28800 0.0539
0.5452 28900 0.0584
0.5471 29000 0.0552
0.5490 29100 0.076
0.5509 29200 0.0861
0.5528 29300 0.067
0.5547 29400 0.0887
0.5565 29500 0.059
0.5584 29600 0.0484
0.5603 29700 0.0703
0.5622 29800 0.0802
0.5641 29900 0.0805
0.5660 30000 0.0737
0.5679 30100 0.0518
0.5697 30200 0.0517
0.5716 30300 0.0806
0.5735 30400 0.0586
0.5754 30500 0.0491
0.5773 30600 0.0591
0.5792 30700 0.066
0.5811 30800 0.0419
0.5830 30900 0.0517
0.5848 31000 0.0539
0.5867 31100 0.0845
0.5886 31200 0.044
0.5905 31300 0.0597
0.5924 31400 0.0556
0.5943 31500 0.0724
0.5962 31600 0.0465
0.5980 31700 0.0585
0.5999 31800 0.0978
0.6018 31900 0.0657
0.6037 32000 0.0438
0.6056 32100 0.0429
0.6075 32200 0.0629
0.6094 32300 0.0591
0.6113 32400 0.0543
0.6131 32500 0.0502
0.6150 32600 0.0733
0.6169 32700 0.0426
0.6188 32800 0.0626
0.6207 32900 0.0406
0.6226 33000 0.0524
0.6245 33100 0.0619
0.6263 33200 0.0633
0.6282 33300 0.0582
0.6301 33400 0.0852
0.6320 33500 0.0482
0.6339 33600 0.0509
0.6358 33700 0.0626
0.6377 33800 0.0609
0.6396 33900 0.0508
0.6414 34000 0.0486
0.6433 34100 0.0508
0.6452 34200 0.0581
0.6471 34300 0.0409
0.6490 34400 0.0703
0.6509 34500 0.0606
0.6528 34600 0.0517
0.6546 34700 0.0493
0.6565 34800 0.0271
0.6584 34900 0.0337
0.6603 35000 0.0369
0.6622 35100 0.0474
0.6641 35200 0.0562
0.6660 35300 0.0663
0.6678 35400 0.0419
0.6697 35500 0.0766
0.6716 35600 0.0439
0.6735 35700 0.0538
0.6754 35800 0.0512
0.6773 35900 0.0388
0.6792 36000 0.0528
0.6811 36100 0.0489
0.6829 36200 0.0454
0.6848 36300 0.0449
0.6867 36400 0.055
0.6886 36500 0.0344
0.6905 36600 0.0485
0.6924 36700 0.0496
0.6943 36800 0.0705
0.6961 36900 0.0617
0.6980 37000 0.054
0.6999 37100 0.0613
0.7018 37200 0.0549
0.7037 37300 0.0378
0.7056 37400 0.0508
0.7075 37500 0.0613
0.7094 37600 0.0602
0.7112 37700 0.0592
0.7131 37800 0.0441
0.7150 37900 0.0445
0.7169 38000 0.0464
0.7188 38100 0.0537
0.7207 38200 0.0521
0.7226 38300 0.0447
0.7244 38400 0.044
0.7263 38500 0.0506
0.7282 38600 0.043
0.7301 38700 0.0441
0.7320 38800 0.0444
0.7339 38900 0.0416
0.7358 39000 0.0556
0.7377 39100 0.0829
0.7395 39200 0.043
0.7414 39300 0.0366
0.7433 39400 0.0457
0.7452 39500 0.0622
0.7471 39600 0.0353
0.7490 39700 0.0597
0.7509 39800 0.0468
0.7527 39900 0.0418
0.7546 40000 0.0606
0.7565 40100 0.0613
0.7584 40200 0.0654
0.7603 40300 0.046
0.7622 40400 0.034
0.7641 40500 0.0378
0.7660 40600 0.0461
0.7678 40700 0.0404
0.7697 40800 0.0583
0.7716 40900 0.0636
0.7735 41000 0.0537
0.7754 41100 0.0336
0.7773 41200 0.0315
0.7792 41300 0.0536
0.7810 41400 0.0532
0.7829 41500 0.0553
0.7848 41600 0.0458
0.7867 41700 0.0372
0.7886 41800 0.0346
0.7905 41900 0.0419
0.7924 42000 0.0461
0.7942 42100 0.0517
0.7961 42200 0.0574
0.7980 42300 0.0411
0.7999 42400 0.0389
0.8018 42500 0.0578
0.8037 42600 0.0637
0.8056 42700 0.0434
0.8075 42800 0.0776
0.8093 42900 0.0644
0.8112 43000 0.0537
0.8131 43100 0.0519
0.8150 43200 0.0241
0.8169 43300 0.0295
0.8188 43400 0.0618
0.8207 43500 0.0275
0.8225 43600 0.0605
0.8244 43700 0.0414
0.8263 43800 0.0446
0.8282 43900 0.0449
0.8301 44000 0.0558
0.8320 44100 0.0336
0.8339 44200 0.0555
0.8358 44300 0.0399
0.8376 44400 0.0319
0.8395 44500 0.0331
0.8414 44600 0.0415
0.8433 44700 0.0424
0.8452 44800 0.0287
0.8471 44900 0.044
0.8490 45000 0.0375
0.8508 45100 0.032
0.8527 45200 0.0406
0.8546 45300 0.0429
0.8565 45400 0.0727
0.8584 45500 0.05
0.8603 45600 0.0436
0.8622 45700 0.0401
0.8641 45800 0.0312
0.8659 45900 0.036
0.8678 46000 0.0558
0.8697 46100 0.0436
0.8716 46200 0.0517
0.8735 46300 0.0361
0.8754 46400 0.038
0.8773 46500 0.0418
0.8791 46600 0.0407
0.8810 46700 0.0336
0.8829 46800 0.0559
0.8848 46900 0.0488
0.8867 47000 0.0463
0.8886 47100 0.0504
0.8905 47200 0.0414
0.8924 47300 0.0428
0.8942 47400 0.0389
0.8961 47500 0.0422
0.8980 47600 0.0533
0.8999 47700 0.0386
0.9018 47800 0.0672
0.9037 47900 0.0505
0.9056 48000 0.0632
0.9074 48100 0.0263
0.9093 48200 0.0448
0.9112 48300 0.0413
0.9131 48400 0.0532
0.9150 48500 0.0503
0.9169 48600 0.0472
0.9188 48700 0.0255
0.9207 48800 0.035
0.9225 48900 0.0353
0.9244 49000 0.0407
0.9263 49100 0.0154
0.9282 49200 0.0535
0.9301 49300 0.0435
0.9320 49400 0.0461
0.9339 49500 0.0288
0.9357 49600 0.0366
0.9376 49700 0.0411
0.9395 49800 0.0605
0.9414 49900 0.0551
0.9433 50000 0.0297
0.9452 50100 0.0388
0.9471 50200 0.0402
0.9489 50300 0.0321
0.9508 50400 0.0538
0.9527 50500 0.036
0.9546 50600 0.0318
0.9565 50700 0.0398
0.9584 50800 0.0405
0.9603 50900 0.0408
0.9622 51000 0.0485
0.9640 51100 0.047
0.9659 51200 0.0452
0.9678 51300 0.0469
0.9697 51400 0.0473
0.9716 51500 0.039
0.9735 51600 0.0579
0.9754 51700 0.0332
0.9772 51800 0.0322
0.9791 51900 0.0324
0.9810 52000 0.035
0.9829 52100 0.0517
0.9848 52200 0.0275
0.9867 52300 0.0466
0.9886 52400 0.0452
0.9905 52500 0.0446
0.9923 52600 0.0357
0.9942 52700 0.0368
0.9961 52800 0.0365
0.9980 52900 0.0303
0.9999 53000 0.0288

Framework Versions

  • Python: 3.11.12
  • Sentence Transformers: 3.4.1
  • Transformers: 4.51.1
  • PyTorch: 2.6.0+cu124
  • Accelerate: 1.5.2
  • Datasets: 3.5.0
  • 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",
}

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}
}