deu05232/LateOn-lr3e-6

PyLate model based on lightonai/LateOn

This is a PyLate model finetuned from lightonai/LateOn on the msmarco-w-instructions dataset. It maps sentences & paragraphs to sequences of 128-dimensional dense vectors and can be used for semantic textual similarity using the MaxSim operator.

Model Details

Model Description

• Model Type: PyLate model
• Base model: lightonai/LateOn
• Document Length: 512 tokens
• Query Length: 512 tokens
• Output Dimensionality: 128 tokens
• Similarity Function: MaxSim
• Training Dataset:
  • msmarco-w-instructions

Model Sources

• Documentation: PyLate Documentation
• Repository: PyLate on GitHub
• Hugging Face: PyLate models on Hugging Face

Full Model Architecture

ColBERT(
  (0): Transformer({'max_seq_length': 511, 'do_lower_case': False, 'architecture': 'ModernBertModel'})
  (1): Dense({'in_features': 768, 'out_features': 1536, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'use_residual': True})
  (2): Dense({'in_features': 1536, 'out_features': 768, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'use_residual': True})
  (3): Dense({'in_features': 768, 'out_features': 128, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'use_residual': False})
)

Usage

First install the PyLate library:

pip install -U pylate

Retrieval

Use this model with PyLate to index and retrieve documents. The index uses FastPLAID for efficient similarity search.

Indexing documents

Load the ColBERT model and initialize the PLAID index, then encode and index your documents:

from pylate import indexes, models, retrieve

# Step 1: Load the ColBERT model
model = models.ColBERT(
    model_name_or_path="pylate_model_id",
)

# Step 2: Initialize the PLAID index
index = indexes.PLAID(
    index_folder="pylate-index",
    index_name="index",
    override=True,  # This overwrites the existing index if any
)

# Step 3: Encode the documents
documents_ids = ["1", "2", "3"]
documents = ["document 1 text", "document 2 text", "document 3 text"]

documents_embeddings = model.encode(
    documents,
    batch_size=32,
    is_query=False,  # Ensure that it is set to False to indicate that these are documents, not queries
    show_progress_bar=True,
)

# Step 4: Add document embeddings to the index by providing embeddings and corresponding ids
index.add_documents(
    documents_ids=documents_ids,
    documents_embeddings=documents_embeddings,
)

Note that you do not have to recreate the index and encode the documents every time. Once you have created an index and added the documents, you can re-use the index later by loading it:

# To load an index, simply instantiate it with the correct folder/name and without overriding it
index = indexes.PLAID(
    index_folder="pylate-index",
    index_name="index",
)

Retrieving top-k documents for queries

Once the documents are indexed, you can retrieve the top-k most relevant documents for a given set of queries. To do so, initialize the ColBERT retriever with the index you want to search in, encode the queries and then retrieve the top-k documents to get the top matches ids and relevance scores:

# Step 1: Initialize the ColBERT retriever
retriever = retrieve.ColBERT(index=index)

# Step 2: Encode the queries
queries_embeddings = model.encode(
    ["query for document 3", "query for document 1"],
    batch_size=32,
    is_query=True,  #  # Ensure that it is set to False to indicate that these are queries
    show_progress_bar=True,
)

# Step 3: Retrieve top-k documents
scores = retriever.retrieve(
    queries_embeddings=queries_embeddings,
    k=10,  # Retrieve the top 10 matches for each query
)

Reranking

If you only want to use the ColBERT model to perform reranking on top of your first-stage retrieval pipeline without building an index, you can simply use rank function and pass the queries and documents to rerank:

from pylate import rank, models

queries = [
    "query A",
    "query B",
]

documents = [
    ["document A", "document B"],
    ["document 1", "document C", "document B"],
]

documents_ids = [
    [1, 2],
    [1, 3, 2],
]

model = models.ColBERT(
    model_name_or_path="pylate_model_id",
)

queries_embeddings = model.encode(
    queries,
    is_query=True,
)

documents_embeddings = model.encode(
    documents,
    is_query=False,
)

reranked_documents = rank.rerank(
    documents_ids=documents_ids,
    queries_embeddings=queries_embeddings,
    documents_embeddings=documents_embeddings,
)

Training Details

Training Dataset

msmarco-w-instructions

• Dataset: msmarco-w-instructions
• Size: 980,250 training samples
• Columns: query, positive, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, negative_9, negative_10, negative_11, negative_12, negative_13, negative_14, and negative_15
• Approximate statistics based on the first 1000 samples:

  	query	positive	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10	negative_11	negative_12	negative_13	negative_14	negative_15
  type	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string	string
  details	min: 5 tokens mean: 67.79 tokens max: 328 tokens	min: 31 tokens mean: 92.53 tokens max: 232 tokens	min: 34 tokens mean: 79.7 tokens max: 256 tokens	min: 20 tokens mean: 78.22 tokens max: 205 tokens	min: 16 tokens mean: 84.03 tokens max: 223 tokens	min: 20 tokens mean: 86.12 tokens max: 256 tokens	min: 20 tokens mean: 85.81 tokens max: 223 tokens	min: 16 tokens mean: 86.63 tokens max: 271 tokens	min: 28 tokens mean: 84.02 tokens max: 227 tokens	min: 20 tokens mean: 87.18 tokens max: 271 tokens	min: 22 tokens mean: 85.9 tokens max: 255 tokens	min: 17 tokens mean: 86.6 tokens max: 244 tokens	min: 22 tokens mean: 87.1 tokens max: 247 tokens	min: 26 tokens mean: 88.25 tokens max: 210 tokens	min: 17 tokens mean: 87.54 tokens max: 244 tokens	min: 25 tokens mean: 85.85 tokens max: 254 tokens	min: 26 tokens mean: 88.0 tokens max: 241 tokens

• Samples:

  query	positive	negative_1	negative_2	negative_3	negative_4	negative_5	negative_6	negative_7	negative_8	negative_9	negative_10	negative_11	negative_12	negative_13	negative_14	negative_15
  how long can you collect unemployment in california	You can get your California EDD Unemployment Benefits Renewed and Extended You can get your California EDD Unemployment Benefits Renewed and Extended. Unemployment benefits in California typically last only 26 weeks after you have been laid off. Most people are so shaken up from their layoff experience, they are not made aware of this till the 26th week of their unemployment benefit period.	How to Calculate Your Unemployment Benefits The amount you receive depends on your weekly earnings prior to being laid-off and on the maximum amount of unemployment benefits paid to each worker. In many states, you will be compensated for half of your earnings, up to a certain maximum. Benefits are usually paid for a maximum of 26 weeks.	Entering the âReal Worldâ If you have been working in a regular job for at least 20 weeks and lose that job - through no fault of your own - you may be eligible for state unemployment benefits. You can file a claim at your local Employment Services Office (listed in your phone book). If you are still out of work when your unemployment benefits run out, you may be eligible for welfare - as long as you still have time left on your five-year limit.	Unemployment Benefits Calculator For more information on Base Period and monetary determination refer unemployment eligibility article. How long will I receive benefits: Usually, most states permit an individual to obtain unemployment for a maximum of 26 weeks, or half the benefit the benefit year. A few states have standardized benefit duration, while most have different durations depending upon the worker.	New York Unemployment Phone Number New York Unemployment Phone Number. Sometimes you just have to talk to a live person to get answers to your questions. In that case, you can call the NY Unemployment Phone Number: Toll Free: 1-888-581-5812. Be prepared for long wait times because lots of people are trying to do the same thing that you are.	Economy Download .xls. The unemployment rate in Danville, California, is 3.20%, with job growth of 0.70%. Future job growth over the next ten years is predicted to be 38.20%. Danville, California Taxes. Danville, California,sales tax rate is 8.50%. Income tax is 6.00%. The income per capita is $61,002, which includes all adults and children.	How to Extend Unemployment Compensation In the United States, federally mandated unemployment compensation typically lasts 26 weeks. However, it may not be possible for you to find work during that time, and you might need extended benefits. During times of high unemployment, states may extend unemployment benefits by a few weeks to a few months. To apply for these benefits, you'll have to do some research and gather the necessary documentation.	Unemployment Rate and Labor Force Unemployment Rate and Labor Force. Each month the Employment Development Department releases revised and preliminary civilian labor force, unemployment rates, and industry employment by geography for California, metropolitan areas, counties, and sub-county areas. The last data released was for June 2017.	California Unemployment Extension If your California unemployment benefits have run out you may be eligible to receive extended unemployment benefits through one of the state or federal unemployment extension programs. This federal unemployment extension programs is called Emergency Unemployment Compensation, 2008 (Also known as EUC08).f you qualify for the EB unemployment extension you will receive additional benefits equal to 20 weeks or 80% of the maximum benefit amount on the original unemployment claim, whichever amount is less.	The Natural Rate, NAIRU, and Monetary Policy Figure 1 shows two additional estimates of the natural rate, together with the actual total civilian unemployment rate and the long-run average rate. The estimate labeled Natural Rate 1 is obtained by removing an estimate of short-run cyclical unemployment from the actual rate.	World Jiu-Jitsu Championship The World Jiu-Jitsu Championship (commonly known as Mundials) is the most prestigious Brazilian Jiu-Jitsu (BJJ) tournament in the world. They are hosted annually by the International Brazilian Jiu-Jitsu Federation. The first Mundial tournament to be held outside of Brazil was in 2007 at California State University, Long Beach in Long Beach, California, United States. This trend continued with the 2008 Mundials, once again in Long Beach, California.	- Claims filed since January 2, 2005 will have a maximum weekly benefit amount of $450. For a listing of eligible entitlements, please see the Unemployment Insurance Benefit Table. You are required to register in California's Job Search assistance website (CalJOBS) within 21 days.	Macroeconomics - 1.) the use of all available resources to produce want-satisfying goods and services; 2.) the situation in which the unemployment rate is equal to the full-employment rate of unemployment and where fractional and structural unemployment occur but not cynical unemployment.	- The Employment Development Department has performed at an unacceptable level for a decade by U.S. Department of Labor standards, according to the California State Auditor. A March 2011 report (and the Labor Department) criticized EDD for not delivering unemployment benefits in a timely fashion.	How Long Can You Collect Unemployment Benefits? How Long Can You Collect Unemployment Benefits? When job losses skyrocketed in 2008, Congress passed emergency measures to supplement state-level unemployment insurance programs, which generally offer six months of benefits. At their peak, the federally backed programs extended benefits to up to 99 weeks in some — though never all — states.	- As of December, 2008, The duration of CO unemployment benefits was 13-26 weeks, but benefits can be extended by Colorado during times of high unemployment or other special circumstances.
  how long can you collect unemployment in california? Unemployment benefits are a vital part of many countries' social safety nets, providing financial assistance to individuals who have lost their jobs. In the United States, each state has its own unemployment insurance program, with varying eligibility requirements and benefit amounts. When searching for information on unemployment benefits, it's essential to consider the specific state and program in question. A relevant document should provide information on the duration of unemployment benefits in a specific state, including any extensions or special programs that may be available. The document should also be written in a clear and concise manner, with a focus on providing practical information to individuals seeking unemployment benefits. A document that simply lists phone numbers or provides general information on unemployment benefits without specific state-level details is not relevant.	duration of unemployment benefits in california In California, individuals can collect unemployment benefits for up to 26 weeks under the regular state program. During periods of high unemployment, additional federal programs may extend this duration. For example, the Pandemic Emergency Unemployment Compensation (PEUC) program provided extensions during the COVID-19 pandemic. It's important to stay informed about any changes or extensions to these programs by regularly checking the California Employment Development Department (EDD) website.	Understanding Unemployment Benefits Unemployment benefits are designed to provide temporary financial assistance to workers who have lost their jobs through no fault of their own. These benefits help individuals cover essential expenses while they search for new employment. Each state in the U.S. administers its own unemployment insurance program, which includes specific eligibility requirements and benefit amounts. It's important to check with your state's unemployment office for detailed information.	Contacting Unemployment Offices If you need assistance with unemployment benefits, you can contact your state's unemployment office. They can provide information on how to apply, eligibility requirements, and benefit amounts. For California, you can reach the Employment Development Department (EDD) at 1-800-300-5616. It's crucial to have all necessary documents ready when you call to expedite the process.	Global Unemployment Benefits Unemployment benefits vary widely across different countries. In some nations, these benefits are more generous and last longer, while in others, they are more limited. For example, in some European countries, unemployment benefits can last up to two years, whereas in other regions, the duration may be much shorter. Understanding the specific policies of each country is essential for those who are unemployed and seeking assistance.	How to Calculate Your Unemployment Benefits The amount you receive depends on your weekly earnings prior to being laid-off and on the maximum amount of unemployment benefits paid to each worker. In many states, you will be compensated for half of your earnings, up to a certain maximum. Benefits are usually paid for a maximum of 26 weeks.	Entering the âReal Worldâ If you have been working in a regular job for at least 20 weeks and lose that job - through no fault of your own - you may be eligible for state unemployment benefits. You can file a claim at your local Employment Services Office (listed in your phone book). If you are still out of work when your unemployment benefits run out, you may be eligible for welfare - as long as you still have time left on your five-year limit.	Unemployment Benefits Calculator For more information on Base Period and monetary determination refer unemployment eligibility article. How long will I receive benefits: Usually, most states permit an individual to obtain unemployment for a maximum of 26 weeks, or half the benefit the benefit year. A few states have standardized benefit duration, while most have different durations depending upon the worker.	New York Unemployment Phone Number New York Unemployment Phone Number. Sometimes you just have to talk to a live person to get answers to your questions. In that case, you can call the NY Unemployment Phone Number: Toll Free: 1-888-581-5812. Be prepared for long wait times because lots of people are trying to do the same thing that you are.	Economy Download .xls. The unemployment rate in Danville, California, is 3.20%, with job growth of 0.70%. Future job growth over the next ten years is predicted to be 38.20%. Danville, California Taxes. Danville, California,sales tax rate is 8.50%. Income tax is 6.00%. The income per capita is $61,002, which includes all adults and children.	How to Extend Unemployment Compensation In the United States, federally mandated unemployment compensation typically lasts 26 weeks. However, it may not be possible for you to find work during that time, and you might need extended benefits. During times of high unemployment, states may extend unemployment benefits by a few weeks to a few months. To apply for these benefits, you'll have to do some research and gather the necessary documentation.	Unemployment Rate and Labor Force Unemployment Rate and Labor Force. Each month the Employment Development Department releases revised and preliminary civilian labor force, unemployment rates, and industry employment by geography for California, metropolitan areas, counties, and sub-county areas. The last data released was for June 2017.	California Unemployment Extension If your California unemployment benefits have run out you may be eligible to receive extended unemployment benefits through one of the state or federal unemployment extension programs. This federal unemployment extension programs is called Emergency Unemployment Compensation, 2008 (Also known as EUC08).f you qualify for the EB unemployment extension you will receive additional benefits equal to 20 weeks or 80% of the maximum benefit amount on the original unemployment claim, whichever amount is less.	The Natural Rate, NAIRU, and Monetary Policy Figure 1 shows two additional estimates of the natural rate, together with the actual total civilian unemployment rate and the long-run average rate. The estimate labeled Natural Rate 1 is obtained by removing an estimate of short-run cyclical unemployment from the actual rate.	World Jiu-Jitsu Championship The World Jiu-Jitsu Championship (commonly known as Mundials) is the most prestigious Brazilian Jiu-Jitsu (BJJ) tournament in the world. They are hosted annually by the International Brazilian Jiu-Jitsu Federation. The first Mundial tournament to be held outside of Brazil was in 2007 at California State University, Long Beach in Long Beach, California, United States. This trend continued with the 2008 Mundials, once again in Long Beach, California.	- Claims filed since January 2, 2005 will have a maximum weekly benefit amount of $450. For a listing of eligible entitlements, please see the Unemployment Insurance Benefit Table. You are required to register in California's Job Search assistance website (CalJOBS) within 21 days.	Macroeconomics - 1.) the use of all available resources to produce want-satisfying goods and services; 2.) the situation in which the unemployment rate is equal to the full-employment rate of unemployment and where fractional and structural unemployment occur but not cynical unemployment.
  who wrote at calvary	Authorship of the hymn 'at calvary' 'At Calvary' is a well-known hymn written by William R. Newell. The hymn reflects on the transformative power of Jesus Christ's sacrifice at Calvary. Newell's lyrics emphasize the themes of redemption and grace, making it a beloved song in many Christian congregations.	- The hymn ap­peared in the At­lant­ic Month­ly in 1862. It was sung at the fun­er­als of Brit­ish states­man Win­ston Church­ill, Amer­i­can sen­at­or Ro­bert Ken­ne­dy, and Am­er­i­can pre­si­dents Ron­ald Rea­gan and Ri­chard Nix­on.	John Donne The lines of these sermons and devotional works would come to influence future works of English literature, such as Ernest Hemingway's For Whom the Bell Tolls, which took its title from a passage in Meditation XVII of Devotions upon Emergent Occasions, and Thomas Merton's No Man is an Island, which took its title from the same source.	The Life of a Song: âIâm a Believerâ Their first single, “Last Train to Clarksville”, written by Tommy Boyce and Bobby Hart, reached number one in the US. Its follow-up, “I’m a Believer”, was the work of a young New York songwriter-for-hire, Neil Diamond, who had initially tried to place it with country singer Eddie Arnold, as well as recording it himself.	Library News - State Government Iric Nathanson wrote about the 1952 and 1956 primaries in a 2008 MinnPost article, Political mischief: Minnesota's 1950s experiment with presidential primaries.. About the repeal, he wrote, The mainly Republican conservatives controlled the state Senate, and they moved first to vote repeal with only minimal debate.	- Practice resurrection. “Manifesto: The Mad Farmer Liberation Front” from The Country of Marriage, Harcourt Brace Jovanovich, Inc. 1973. Also published by Counterpoint Press in The Selected Poems of Wendell Berry, 1999; The Mad Farmer Poems, 2008; New Collected Poems, 2012.	Alan Menken According to Alan Menken ... after age 30 it is a downhill plunge. Menken wrote several shows that were not produced, including Atina, Evil Queen of the Galaxy (1980), with lyrics by Steve Brown. He also wrote The Thorn with lyrics by Brown, which was commissioned by Divine in 1980.	Henry Wadsworth Longfellow Poems such as “Paul Revere’s Ride,” Evangeline, A Tale of Acadie (1847), and “A Psalm of Life” became mainstays of national culture, long remembered by generations of readers who studied them in school.	Do Not Stand at My Grave and Weep Do Not Stand at My Grave and Weep. Do Not Stand at My Grave and Weep is a poem written in 1932 by Mary Elizabeth Frye. Although the origin of the poem was disputed until later in her life, Mary Frye's authorship was confirmed in 1998 after research by Abigail Van Buren, a newspaper columnist. 1 Origins.	- Randall's most famous poem is The Ballad of Birmingham, written during the 1960s, about the 1963 bombing of the church Martin Luther King, Jr. belonged to in Birmingham, Alabama.[2] Randall's poetry	Best Famous Romantic Poets of All-time Andreas Kalvos. Andreas Kalvos (Greek: ἀνδρέας κάλβος, also transliterated as Andreas Calvos; 1 April 1792 – November 3, 1869) was a Greek poet of the Romantic school. He published only two collections of poems — the Lyra of 1824 and the Lyrica of 1826. He was a contemporary of the poets Ugo Foscolo and Dionysios Solomos.	The Hollywood Ten This is a static page that is no longer maintained. The Hollywood Ten. American novelist and journalist Alvah Bessie wrote screenplays for Warner Brothers and other studios during the mid and late '40s. As a screenwriter, he was nominated for an Academy Award for Best Original Story for the patriotic Warner's film Objective Burma(1945).	Austin Miles' Methodist Hymn: In The Garden You Are At: AllSands Home > Religious > Christian > Austin Miles' Methodist hymn: in the garden. Written in 1912 by C. Austin Miles, as a commission by a music publisher, “In the Garden” was no less spiritually inspired for all its secular beginnings. Miles left an accounting of the writing of it. He describes sitting in a darkened room and taking up the Bible for inspiration.	- He also wrote a paper concerning aggression instinct and Freud did not approve. In addition, Alfred wrote a paper on children's feelings of inferiority. Which agreed with Freud's sexual notions. Freud named Adler president of Viennese Analytic Society co-editor of organization newsletter.	Sermons on Mark - Brian Bill Browse Collections. The Roman Centurion. Mark 15:39 tells us that the Roman centurion, who had witnessed the suffering and death of Christ at Calvary first-hand, said: Truly this man was the Son of God!.	Alan Menken According to Alan Menken ... after age 30 it is a downhill plunge. Menken wrote several shows that were not produced, including Atina, Evil Queen of the Galaxy (1980), with lyrics by Steve Brown. He also wrote The Thorn with lyrics by Brown, which was commissioned by Divine in 1980.

• Loss: main.TrimmedCachedContrastive

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 64
• num_train_epochs: 1.0
• learning_rate: 3e-06
• bf16: True
• dataloader_num_workers: 4

All Hyperparameters

Click to expand

• per_device_train_batch_size: 64
• num_train_epochs: 1.0
• max_steps: -1
• learning_rate: 3e-06
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_steps: 0
• optim: adamw_torch_fused
• optim_args: None
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• optim_target_modules: None
• gradient_accumulation_steps: 1
• average_tokens_across_devices: True
• max_grad_norm: 1.0
• label_smoothing_factor: 0.0
• bf16: True
• fp16: False
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• use_liger_kernel: False
• liger_kernel_config: None
• use_cache: False
• neftune_noise_alpha: None
• torch_empty_cache_steps: None
• auto_find_batch_size: False
• log_on_each_node: True
• logging_nan_inf_filter: True
• include_num_input_tokens_seen: no
• log_level: passive
• log_level_replica: warning
• disable_tqdm: False
• project: huggingface
• trackio_space_id: trackio
• eval_strategy: no
• per_device_eval_batch_size: 8
• prediction_loss_only: True
• eval_on_start: False
• eval_do_concat_batches: True
• eval_use_gather_object: False
• eval_accumulation_steps: None
• include_for_metrics: []
• batch_eval_metrics: False
• save_only_model: False
• save_on_each_node: False
• enable_jit_checkpoint: False
• push_to_hub: False
• hub_private_repo: None
• hub_model_id: None
• hub_strategy: every_save
• hub_always_push: False
• hub_revision: None
• load_best_model_at_end: False
• ignore_data_skip: False
• restore_callback_states_from_checkpoint: False
• full_determinism: False
• seed: 42
• data_seed: None
• use_cpu: False
• 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
• dataloader_drop_last: True
• dataloader_num_workers: 4
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• dataloader_prefetch_factor: None
• remove_unused_columns: True
• label_names: None
• train_sampling_strategy: random
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• ddp_backend: None
• ddp_timeout: 1800
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• deepspeed: None
• debug: []
• skip_memory_metrics: True
• do_predict: False
• resume_from_checkpoint: None
• warmup_ratio: None
• local_rank: -1
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss
0.0013	10	4.1241
0.0026	20	2.8371
0.0039	30	2.2661
0.0052	40	1.9584
0.0065	50	1.7307
0.0078	60	1.4898
0.0091	70	1.5371
0.0104	80	1.3562
0.0118	90	1.4236
0.0131	100	1.3141
0.0144	110	1.3822
0.0157	120	1.4062
0.0170	130	1.3419
0.0183	140	1.2586
0.0196	150	1.2167
0.0209	160	1.2426
0.0222	170	1.1807
0.0235	180	1.2476
0.0248	190	1.1818
0.0261	200	1.1478
0.0274	210	1.2303
0.0287	220	1.0560
0.0300	230	1.0606
0.0313	240	1.1303
0.0326	250	1.0256
0.0340	260	1.1614
0.0353	270	1.0904
0.0366	280	1.1257
0.0379	290	1.1255
0.0392	300	0.9967
0.0405	310	1.1079
0.0418	320	0.9494
0.0431	330	1.0317
0.0444	340	1.0438
0.0457	350	1.1570
0.0470	360	1.1040
0.0483	370	0.9556
0.0496	380	1.0233
0.0509	390	1.0484
0.0522	400	0.9850
0.0535	410	1.0004
0.0548	420	0.9260
0.0562	430	0.8822
0.0575	440	1.0373
0.0588	450	0.9045
0.0601	460	0.9923
0.0614	470	1.0748
0.0627	480	1.0669
0.0640	490	0.8912
0.0653	500	0.9612
0.0666	510	0.9882
0.0679	520	1.0082
0.0692	530	0.9944
0.0705	540	0.9140
0.0718	550	0.9537
0.0731	560	1.0428
0.0744	570	0.9087
0.0757	580	0.8410
0.0770	590	0.9253
0.0783	600	0.9441
0.0797	610	1.0326
0.0810	620	0.9747
0.0823	630	0.9971
0.0836	640	0.8355
0.0849	650	0.9123
0.0862	660	1.0012
0.0875	670	1.0798
0.0888	680	0.9337
0.0901	690	0.9997
0.0914	700	0.8419
0.0927	710	0.9200
0.0940	720	0.8775
0.0953	730	0.9584
0.0966	740	0.9330
0.0979	750	0.8711
0.0992	760	0.8229
0.1005	770	0.8895
0.1019	780	0.8803
0.1032	790	0.9152
0.1045	800	0.9143
0.1058	810	0.9482
0.1071	820	0.9758
0.1084	830	0.9700
0.1097	840	1.0181
0.1110	850	0.9483
0.1123	860	0.8776
0.1136	870	0.9238
0.1149	880	0.9389
0.1162	890	0.8571
0.1175	900	0.8621
0.1188	910	0.8697
0.1201	920	0.8729
0.1214	930	0.9384
0.1227	940	0.9007
0.1241	950	0.9396
0.1254	960	0.8363
0.1267	970	0.9272
0.1280	980	0.8560
0.1293	990	0.8877
0.1306	1000	0.9212
0.1319	1010	0.9819
0.1332	1020	0.9179
0.1345	1030	0.9129
0.1358	1040	0.9637
0.1371	1050	0.9070
0.1384	1060	0.8725
0.1397	1070	0.8731
0.1410	1080	0.8400
0.1423	1090	0.9365
0.1436	1100	0.7337
0.1449	1110	0.8697
0.1463	1120	0.8485
0.1476	1130	0.8814
0.1489	1140	0.8570
0.1502	1150	0.8939
0.1515	1160	0.9129
0.1528	1170	0.8892
0.1541	1180	0.8930
0.1554	1190	0.8354
0.1567	1200	0.8634
0.1580	1210	0.9217
0.1593	1220	0.8720
0.1606	1230	0.8493
0.1619	1240	0.8742
0.1632	1250	0.9072
0.1645	1260	0.7950
0.1658	1270	0.8887
0.1671	1280	0.8406
0.1685	1290	0.9094
0.1698	1300	0.8946
0.1711	1310	0.8298
0.1724	1320	0.8856
0.1737	1330	0.8995
0.1750	1340	0.8454
0.1763	1350	0.8306
0.1776	1360	0.8278
0.1789	1370	0.8478
0.1802	1380	0.8227
0.1815	1390	1.0017
0.1828	1400	0.8272
0.1841	1410	0.8474
0.1854	1420	0.9806
0.1867	1430	0.8341
0.1880	1440	0.8905
0.1893	1450	0.8962
0.1907	1460	0.8871
0.1920	1470	0.8695
0.1933	1480	0.8829
0.1946	1490	0.8736
0.1959	1500	0.8346
0.1972	1510	0.8285
0.1985	1520	0.8356
0.1998	1530	0.8031
0.2011	1540	0.8558
0.2024	1550	0.9077
0.2037	1560	0.8492
0.2050	1570	0.9099
0.2063	1580	0.8730
0.2076	1590	0.7533
0.2089	1600	0.8103
0.2102	1610	0.7879
0.2115	1620	0.7881
0.2128	1630	0.8609
0.2142	1640	0.9359
0.2155	1650	0.8915
0.2168	1660	0.9291
0.2181	1670	0.8826
0.2194	1680	0.8121
0.2207	1690	0.8621
0.2220	1700	0.8247
0.2233	1710	0.8911
0.2246	1720	0.8772
0.2259	1730	0.7973
0.2272	1740	0.7767
0.2285	1750	0.8695
0.2298	1760	0.9139
0.2311	1770	0.7866
0.2324	1780	0.7210
0.2337	1790	0.7892
0.2350	1800	0.7829
0.2364	1810	0.8066
0.2377	1820	0.8346
0.2390	1830	0.8499
0.2403	1840	0.7384
0.2416	1850	0.8109
0.2429	1860	0.9094
0.2442	1870	0.8888
0.2455	1880	0.7784
0.2468	1890	0.8248
0.2481	1900	0.8234
0.2494	1910	0.8888
0.2507	1920	0.9432
0.2520	1930	0.8806
0.2533	1940	0.8047
0.2546	1950	0.8460
0.2559	1960	0.8977
0.2572	1970	0.7974
0.2586	1980	0.8733
0.2599	1990	0.8327
0.2612	2000	0.8773
0.2625	2010	0.7844
0.2638	2020	0.8359
0.2651	2030	0.8635
0.2664	2040	0.7958
0.2677	2050	0.8180
0.2690	2060	0.8772
0.2703	2070	0.8374
0.2716	2080	0.8788
0.2729	2090	0.8746
0.2742	2100	0.8516
0.2755	2110	0.8607
0.2768	2120	0.9029
0.2781	2130	0.7731
0.2794	2140	0.8635
0.2808	2150	0.8676
0.2821	2160	0.7578
0.2834	2170	0.8372
0.2847	2180	0.8645
0.2860	2190	0.8326
0.2873	2200	0.9053
0.2886	2210	0.8490
0.2899	2220	0.7719
0.2912	2230	0.8440
0.2925	2240	0.8656
0.2938	2250	0.7800
0.2951	2260	0.8757
0.2964	2270	0.8617
0.2977	2280	0.8543
0.2990	2290	0.8338
0.3003	2300	0.8812
0.3016	2310	0.8055
0.3030	2320	0.7929
0.3043	2330	0.8201
0.3056	2340	0.8271
0.3069	2350	0.8905
0.3082	2360	0.8436
0.3095	2370	0.8669
0.3108	2380	0.7828
0.3121	2390	0.7620
0.3134	2400	0.8527
0.3147	2410	0.8083
0.3160	2420	0.8126
0.3173	2430	0.8101
0.3186	2440	0.8388
0.3199	2450	0.8895
0.3212	2460	0.8458
0.3225	2470	0.8054
0.3238	2480	0.7221
0.3252	2490	0.7809
0.3265	2500	0.7915
0.3278	2510	0.8591
0.3291	2520	0.8021
0.3304	2530	0.7425
0.3317	2540	0.8247
0.3330	2550	0.8435
0.3343	2560	0.7866
0.3356	2570	0.7878
0.3369	2580	0.8317
0.3382	2590	0.7522
0.3395	2600	0.8039
0.3408	2610	0.8489
0.3421	2620	0.7299
0.3434	2630	0.8043
0.3447	2640	0.8395
0.3460	2650	0.7734
0.3473	2660	0.7523
0.3487	2670	0.8702
0.3500	2680	0.8239
0.3513	2690	0.7868
0.3526	2700	0.8003
0.3539	2710	0.8488
0.3552	2720	0.7707
0.3565	2730	0.8754
0.3578	2740	0.8348
0.3591	2750	0.7345
0.3604	2760	0.7628
0.3617	2770	0.7703
0.3630	2780	0.8843
0.3643	2790	0.7772
0.3656	2800	0.7427
0.3669	2810	0.8406
0.3682	2820	0.9360
0.3695	2830	0.8008
0.3709	2840	0.7446
0.3722	2850	0.8064
0.3735	2860	0.7837
0.3748	2870	0.7351
0.3761	2880	0.8952
0.3774	2890	0.8566
0.3787	2900	0.7715
0.3800	2910	0.8267
0.3813	2920	0.9310
0.3826	2930	0.8935
0.3839	2940	0.7759
0.3852	2950	0.8701
0.3865	2960	0.7168
0.3878	2970	0.8252
0.3891	2980	0.7455
0.3904	2990	0.8772
0.3917	3000	0.8928
0.3931	3010	0.8285
0.3944	3020	0.7222
0.3957	3030	0.8113
0.3970	3040	0.7928
0.3983	3050	0.7186
0.3996	3060	0.8435
0.4009	3070	0.7230
0.4022	3080	0.8045
0.4035	3090	0.7728
0.4048	3100	0.8131
0.4061	3110	0.8285
0.4074	3120	0.7349
0.4087	3130	0.8085
0.4100	3140	0.7102
0.4113	3150	0.8203
0.4126	3160	0.8436
0.4139	3170	0.8591
0.4153	3180	0.7046
0.4166	3190	0.8201
0.4179	3200	0.7433
0.4192	3210	0.7592
0.4205	3220	0.7605
0.4218	3230	0.8125
0.4231	3240	0.8853
0.4244	3250	0.7063
0.4257	3260	0.8202
0.4270	3270	0.7601
0.4283	3280	0.7559
0.4296	3290	0.7856
0.4309	3300	0.7569
0.4322	3310	0.7500
0.4335	3320	0.8080
0.4348	3330	0.8428
0.4361	3340	0.8172
0.4375	3350	0.8979
0.4388	3360	0.7550
0.4401	3370	0.7687
0.4414	3380	0.7693
0.4427	3390	0.7742
0.4440	3400	0.8252
0.4453	3410	0.7875
0.4466	3420	0.7620
0.4479	3430	0.8442
0.4492	3440	0.7814
0.4505	3450	0.7365
0.4518	3460	0.8711
0.4531	3470	0.8290
0.4544	3480	0.7543
0.4557	3490	0.7527
0.4570	3500	0.6942
0.4583	3510	0.7181
0.4597	3520	0.8150
0.4610	3530	0.8321
0.4623	3540	0.8188
0.4636	3550	0.8157
0.4649	3560	0.8073
0.4662	3570	0.8673
0.4675	3580	0.6867
0.4688	3590	0.8391
0.4701	3600	0.8145
0.4714	3610	0.8513
0.4727	3620	0.7712
0.4740	3630	0.7962
0.4753	3640	0.8957
0.4766	3650	0.8957
0.4779	3660	0.6481
0.4792	3670	0.7931
0.4805	3680	0.7991
0.4818	3690	0.7933
0.4832	3700	0.9388
0.4845	3710	0.7060
0.4858	3720	0.7466
0.4871	3730	0.7238
0.4884	3740	0.7130
0.4897	3750	0.8474
0.4910	3760	0.7724
0.4923	3770	0.7900
0.4936	3780	0.7720
0.4949	3790	0.8877
0.4962	3800	0.7229
0.4975	3810	0.6991
0.4988	3820	0.7279
0.5001	3830	0.7960
0.5014	3840	0.7579
0.5027	3850	0.8374
0.5040	3860	0.7661
0.5054	3870	0.8268
0.5067	3880	0.7361
0.5080	3890	0.7805
0.5093	3900	0.7590
0.5106	3910	0.6605
0.5119	3920	0.7997
0.5132	3930	0.8363
0.5145	3940	0.7689
0.5158	3950	0.8032
0.5171	3960	0.7550
0.5184	3970	0.7496
0.5197	3980	0.7701
0.5210	3990	0.7158
0.5223	4000	0.7513
0.5236	4010	0.7963
0.5249	4020	0.8003
0.5262	4030	0.7688
0.5276	4040	0.8495
0.5289	4050	0.7667
0.5302	4060	0.8569
0.5315	4070	0.8080
0.5328	4080	0.8531
0.5341	4090	0.8238
0.5354	4100	0.8626
0.5367	4110	0.8411
0.5380	4120	0.7807
0.5393	4130	0.6896
0.5406	4140	0.7627
0.5419	4150	0.8739
0.5432	4160	0.7287
0.5445	4170	0.7169
0.5458	4180	0.8312
0.5471	4190	0.7652
0.5484	4200	0.7286
0.5498	4210	0.7924
0.5511	4220	0.7778
0.5524	4230	0.8103
0.5537	4240	0.7886
0.5550	4250	0.7644
0.5563	4260	0.7547
0.5576	4270	0.7798
0.5589	4280	0.8683
0.5602	4290	0.7257
0.5615	4300	0.7857
0.5628	4310	0.8129
0.5641	4320	0.7831
0.5654	4330	0.7877
0.5667	4340	0.7680
0.5680	4350	0.7091
0.5693	4360	0.7575
0.5706	4370	0.7583
0.5720	4380	0.8193
0.5733	4390	0.8437
0.5746	4400	0.8347
0.5759	4410	0.7752
0.5772	4420	0.7620
0.5785	4430	0.7751
0.5798	4440	0.8726
0.5811	4450	0.7828
0.5824	4460	0.8385
0.5837	4470	0.9128
0.5850	4480	0.7595
0.5863	4490	0.8009
0.5876	4500	0.8424
0.5889	4510	0.7377
0.5902	4520	0.8613
0.5915	4530	0.7622
0.5928	4540	0.8207
0.5941	4550	0.7604
0.5955	4560	0.8750
0.5968	4570	0.8294
0.5981	4580	0.7347
0.5994	4590	0.7975
0.6007	4600	0.7597
0.6020	4610	0.7636
0.6033	4620	0.7509
0.6046	4630	0.7530
0.6059	4640	0.7690
0.6072	4650	0.7803
0.6085	4660	0.7515
0.6098	4670	0.7762
0.6111	4680	0.7502
0.6124	4690	0.7482
0.6137	4700	0.8297
0.6150	4710	0.7898
0.6163	4720	0.7356
0.6177	4730	0.7574
0.6190	4740	0.8057
0.6203	4750	0.7569
0.6216	4760	0.7378
0.6229	4770	0.8152
0.6242	4780	0.7709
0.6255	4790	0.7604
0.6268	4800	0.7461
0.6281	4810	0.7529
0.6294	4820	0.7480
0.6307	4830	0.7917
0.6320	4840	0.7269
0.6333	4850	0.8414
0.6346	4860	0.8499
0.6359	4870	0.7135
0.6372	4880	0.7623
0.6385	4890	0.8378
0.6399	4900	0.7733
0.6412	4910	0.9060
0.6425	4920	0.7689
0.6438	4930	0.6684
0.6451	4940	0.7483
0.6464	4950	0.8208
0.6477	4960	0.7992
0.6490	4970	0.7991
0.6503	4980	0.8211
0.6516	4990	0.7736
0.6529	5000	0.9109
0.6542	5010	0.7075
0.6555	5020	0.7292
0.6568	5030	0.7476
0.6581	5040	0.7289
0.6594	5050	0.7232
0.6607	5060	0.8342
0.6621	5070	0.7669
0.6634	5080	0.7609
0.6647	5090	0.8202
0.6660	5100	0.8032
0.6673	5110	0.7540
0.6686	5120	0.8100
0.6699	5130	0.6942
0.6712	5140	0.7320
0.6725	5150	0.7146
0.6738	5160	0.7670
0.6751	5170	0.8800
0.6764	5180	0.7697
0.6777	5190	0.8039
0.6790	5200	0.7995
0.6803	5210	0.7314
0.6816	5220	0.7594
0.6829	5230	0.7999
0.6843	5240	0.7533
0.6856	5250	0.7092
0.6869	5260	0.7466
0.6882	5270	0.6936
0.6895	5280	0.7402
0.6908	5290	0.7594
0.6921	5300	0.7587
0.6934	5310	0.7959
0.6947	5320	0.8297
0.6960	5330	0.8336
0.6973	5340	0.7205
0.6986	5350	0.6937
0.6999	5360	0.8041
0.7012	5370	0.7659
0.7025	5380	0.7682
0.7038	5390	0.7360
0.7051	5400	0.7717
0.7065	5410	0.8888
0.7078	5420	0.8114
0.7091	5430	0.8217
0.7104	5440	0.7287
0.7117	5450	0.7950
0.7130	5460	0.8222
0.7143	5470	0.7539
0.7156	5480	0.7269
0.7169	5490	0.7167
0.7182	5500	0.7383
0.7195	5510	0.8223
0.7208	5520	0.8210
0.7221	5530	0.7702
0.7234	5540	0.7866
0.7247	5550	0.7491
0.7260	5560	0.7701
0.7273	5570	0.7114
0.7286	5580	0.7916
0.7300	5590	0.7682
0.7313	5600	0.8354
0.7326	5610	0.7622
0.7339	5620	0.8125
0.7352	5630	0.7220
0.7365	5640	0.7405
0.7378	5650	0.7980
0.7391	5660	0.7745
0.7404	5670	0.9737
0.7417	5680	0.7769
0.7430	5690	0.8620
0.7443	5700	0.6889
0.7456	5710	0.8083
0.7469	5720	0.7902
0.7482	5730	0.8222
0.7495	5740	0.8088
0.7508	5750	0.8075
0.7522	5760	0.6921
0.7535	5770	0.7284
0.7548	5780	0.8867
0.7561	5790	0.7243
0.7574	5800	0.7313
0.7587	5810	0.8139
0.7600	5820	0.7935
0.7613	5830	0.8095
0.7626	5840	0.7938
0.7639	5850	0.7776
0.7652	5860	0.8378
0.7665	5870	0.7470
0.7678	5880	0.7562
0.7691	5890	0.6729
0.7704	5900	0.7178
0.7717	5910	0.7771
0.7730	5920	0.7328
0.7744	5930	0.7740
0.7757	5940	0.8560
0.7770	5950	0.7739
0.7783	5960	0.7762
0.7796	5970	0.8079
0.7809	5980	0.8013
0.7822	5990	0.8442
0.7835	6000	0.7409
0.7848	6010	0.7834
0.7861	6020	0.7137
0.7874	6030	0.8295
0.7887	6040	0.7498
0.7900	6050	0.7775
0.7913	6060	0.7186
0.7926	6070	0.7373
0.7939	6080	0.7400
0.7952	6090	0.7301
0.7966	6100	0.7360
0.7979	6110	0.8327
0.7992	6120	0.7507
0.8005	6130	0.7760
0.8018	6140	0.7588
0.8031	6150	0.8078
0.8044	6160	0.7482
0.8057	6170	0.6471
0.8070	6180	0.7695
0.8083	6190	0.7990
0.8096	6200	0.7473
0.8109	6210	0.7527
0.8122	6220	0.8181
0.8135	6230	0.8078
0.8148	6240	0.7956
0.8161	6250	0.7407
0.8174	6260	0.7082
0.8188	6270	0.8211
0.8201	6280	0.7198
0.8214	6290	0.7233
0.8227	6300	0.7720
0.8240	6310	0.7409
0.8253	6320	0.7167
0.8266	6330	0.7577
0.8279	6340	0.8237
0.8292	6350	0.7970
0.8305	6360	0.7698
0.8318	6370	0.8064
0.8331	6380	0.7851
0.8344	6390	0.7445
0.8357	6400	0.7138
0.8370	6410	0.7981
0.8383	6420	0.7262
0.8396	6430	0.7547
0.8410	6440	0.7476
0.8423	6450	0.7965
0.8436	6460	0.7502
0.8449	6470	0.7208
0.8462	6480	0.8069
0.8475	6490	0.8084
0.8488	6500	0.7912
0.8501	6510	0.7506
0.8514	6520	0.8591
0.8527	6530	0.8110
0.8540	6540	0.7838
0.8553	6550	0.7599
0.8566	6560	0.7882
0.8579	6570	0.6755
0.8592	6580	0.7171
0.8605	6590	0.7401
0.8618	6600	0.7599
0.8631	6610	0.8081
0.8645	6620	0.6546
0.8658	6630	0.6723
0.8671	6640	0.8204
0.8684	6650	0.8281
0.8697	6660	0.7965
0.8710	6670	0.7267
0.8723	6680	0.8046
0.8736	6690	0.7643
0.8749	6700	0.7924
0.8762	6710	0.7930
0.8775	6720	0.8174
0.8788	6730	0.7166
0.8801	6740	0.7288
0.8814	6750	0.6725
0.8827	6760	0.7737
0.8840	6770	0.8376
0.8853	6780	0.7470
0.8867	6790	0.8291
0.8880	6800	0.7724
0.8893	6810	0.7442
0.8906	6820	0.8279
0.8919	6830	0.7364
0.8932	6840	0.6710
0.8945	6850	0.8329
0.8958	6860	0.8047
0.8971	6870	0.8254
0.8984	6880	0.7459
0.8997	6890	0.8340
0.9010	6900	0.7753
0.9023	6910	0.7980
0.9036	6920	0.7962
0.9049	6930	0.7187
0.9062	6940	0.7745
0.9075	6950	0.7620
0.9089	6960	0.7682
0.9102	6970	0.8309
0.9115	6980	0.8326
0.9128	6990	0.7608
0.9141	7000	0.7876
0.9154	7010	0.7838
0.9167	7020	0.7493
0.9180	7030	0.7579
0.9193	7040	0.7321
0.9206	7050	0.7863
0.9219	7060	0.8641
0.9232	7070	0.6685
0.9245	7080	0.7906
0.9258	7090	0.7406
0.9271	7100	0.8443
0.9284	7110	0.8147
0.9297	7120	0.8147
0.9311	7130	0.7796
0.9324	7140	0.7589
0.9337	7150	0.7080
0.9350	7160	0.7769
0.9363	7170	0.6949
0.9376	7180	0.8385
0.9389	7190	0.7781
0.9402	7200	0.8011
0.9415	7210	0.7343
0.9428	7220	0.7749
0.9441	7230	0.7174
0.9454	7240	0.7304
0.9467	7250	0.7701
0.9480	7260	0.8001
0.9493	7270	0.7392
0.9506	7280	0.8152
0.9519	7290	0.8281
0.9533	7300	0.7433
0.9546	7310	0.8082
0.9559	7320	0.6789
0.9572	7330	0.7545
0.9585	7340	0.7387
0.9598	7350	0.7491
0.9611	7360	0.7295
0.9624	7370	0.7872
0.9637	7380	0.8013
0.9650	7390	0.6952
0.9663	7400	0.7472
0.9676	7410	0.8022
0.9689	7420	0.7600
0.9702	7430	0.6862
0.9715	7440	0.7874
0.9728	7450	0.7317
0.9741	7460	0.7049
0.9755	7470	0.7987
0.9768	7480	0.8687
0.9781	7490	0.7509
0.9794	7500	0.7795
0.9807	7510	0.6646
0.9820	7520	0.7784
0.9833	7530	0.8844
0.9846	7540	0.8417
0.9859	7550	0.6884
0.9872	7560	0.8561
0.9885	7570	0.8252
0.9898	7580	0.7963
0.9911	7590	0.7191
0.9924	7600	0.8011
0.9937	7610	0.7120
0.9950	7620	0.7097
0.9963	7630	0.8293
0.9976	7640	0.7789
0.9990	7650	0.7847

Framework Versions

• Python: 3.11.17
• Sentence Transformers: 5.3.0
• PyLate: 1.6.0
• Transformers: 5.3.0
• PyTorch: 2.8.0+cu128
• Accelerate: 1.15.0
• Datasets: 5.1.0
• Tokenizers: 0.22.2

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"
}

PyLate

@inproceedings{DBLP:conf/cikm/ChaffinS25,
  author       = {Antoine Chaffin and
                  Rapha{"{e}}l Sourty},
  editor       = {Meeyoung Cha and
                  Chanyoung Park and
                  Noseong Park and
                  Carl Yang and
                  Senjuti Basu Roy and
                  Jessie Li and
                  Jaap Kamps and
                  Kijung Shin and
                  Bryan Hooi and
                  Lifang He},
  title        = {PyLate: Flexible Training and Retrieval for Late Interaction Models},
  booktitle    = {Proceedings of the 34th {ACM} International Conference on Information
                  and Knowledge Management, {CIKM} 2025, Seoul, Republic of Korea, November
                  10-14, 2025},
  pages        = {6334--6339},
  publisher    = {{ACM}},
  year         = {2025},
  url          = {https://github.com/lightonai/pylate},
  doi          = {10.1145/3746252.3761608},
}

TrimmedCachedContrastive

@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}