bansalaman18/reranker-msmarco-v1.1-ettin-encoder-68m-listnet

CrossEncoder based on jhu-clsp/ettin-encoder-68m

This is a Cross Encoder model finetuned from jhu-clsp/ettin-encoder-68m on the ms_marco dataset using the sentence-transformers library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.

Model Details

Model Description

• Model Type: Cross Encoder
• Base model: jhu-clsp/ettin-encoder-68m
• Maximum Sequence Length: 7999 tokens
• Number of Output Labels: 1 label
• Training Dataset:
  • ms_marco
• Language: en

Model Sources

• Documentation: Sentence Transformers Documentation
• Documentation: Cross Encoder Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Cross Encoders on Hugging Face

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("bansalaman18/reranker-msmarco-v1.1-ettin-encoder-68m-listnet")
# Get scores for pairs of texts
pairs = [
    ['what is the name of the vaccine for MMR', 'Measles, mumps and rubella (MMR) vaccines. The MMR vaccine also comes in combination with chickenpox (MMRV) for 18 month old children and contains small amounts of each of the viruses at a reduced strength and a small amount of the antibiotic neomycin. MMR is given at four years of age to children who did not get their second MMR vaccine at 18 months of age. The four year old MMR dose ends in December 2015. All people born during or since 1966 should check their immunisation status to ensure they have had two doses of a measles containing vaccine.'],
    ['what is the name of the vaccine for MMR', 'The MMR vaccine protects children against all three diseases and is given at 12 months of age. A second dose using MMRV vaccine is given at 18 months of age to also protect children from chickenpox. MMR is given at four years of age to children who did not get their second MMR vaccine at 18 months of age. The four year old MMR dose ends in December 2015. All people born during or since 1966 should check their immunisation status to ensure they have had two doses of a measles containing vaccine.'],
    ['what is the name of the vaccine for MMR', 'Measles vaccine is a vaccine that is very effective at preventing measles. After one dose 85% of children nine months of age and 95% over twelve months of age are immune. Nearly all of those who do not develop immunity after a single dose develop it after a second dose. Measles mumps rubella vaccine (MMR-II); MMR vaccine is a live attenuated viral vaccine used to induce immunity against measles, mumps and rubella.'],
    ['what is the name of the vaccine for MMR', 'The name of the measles vaccination is MMR. This is a three-in-one vaccination to protect against measles, mumps and rubella. The measles vaccine is a living, weakened form of the natural  measles virus. To make certain vaccines, viruses are weakened by  a process called cell culture adaptatio … n.. Cell culture  adaptation changes the natural measles virus so that it behaves  differently once it enters the body.'],
    ['what is the name of the vaccine for MMR', 'When the first dose of measles, mumps, rubella, and varicella vaccines is administered at ages 48 months and older, use of MMRV vaccine generally is preferred over separate injections of MMR and varicella vaccines. However, the overall risk of febrile seizures is very low for both options (about 8 out of every 10,000 children vaccinated with MMRV vaccine when they are 12-23 months old, and about 4 out of every 10,000 children vaccinated with the MMR and varicella vaccines at the same visit when they are 12-23 months old).'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'what is the name of the vaccine for MMR',
    [
        'Measles, mumps and rubella (MMR) vaccines. The MMR vaccine also comes in combination with chickenpox (MMRV) for 18 month old children and contains small amounts of each of the viruses at a reduced strength and a small amount of the antibiotic neomycin. MMR is given at four years of age to children who did not get their second MMR vaccine at 18 months of age. The four year old MMR dose ends in December 2015. All people born during or since 1966 should check their immunisation status to ensure they have had two doses of a measles containing vaccine.',
        'The MMR vaccine protects children against all three diseases and is given at 12 months of age. A second dose using MMRV vaccine is given at 18 months of age to also protect children from chickenpox. MMR is given at four years of age to children who did not get their second MMR vaccine at 18 months of age. The four year old MMR dose ends in December 2015. All people born during or since 1966 should check their immunisation status to ensure they have had two doses of a measles containing vaccine.',
        'Measles vaccine is a vaccine that is very effective at preventing measles. After one dose 85% of children nine months of age and 95% over twelve months of age are immune. Nearly all of those who do not develop immunity after a single dose develop it after a second dose. Measles mumps rubella vaccine (MMR-II); MMR vaccine is a live attenuated viral vaccine used to induce immunity against measles, mumps and rubella.',
        'The name of the measles vaccination is MMR. This is a three-in-one vaccination to protect against measles, mumps and rubella. The measles vaccine is a living, weakened form of the natural  measles virus. To make certain vaccines, viruses are weakened by  a process called cell culture adaptatio … n.. Cell culture  adaptation changes the natural measles virus so that it behaves  differently once it enters the body.',
        'When the first dose of measles, mumps, rubella, and varicella vaccines is administered at ages 48 months and older, use of MMRV vaccine generally is preferred over separate injections of MMR and varicella vaccines. However, the overall risk of febrile seizures is very low for both options (about 8 out of every 10,000 children vaccinated with MMRV vaccine when they are 12-23 months old, and about 4 out of every 10,000 children vaccinated with the MMR and varicella vaccines at the same visit when they are 12-23 months old).',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Training Details

Training Dataset

ms_marco

• Dataset: ms_marco at a47ee7a
• Size: 78,704 training samples
• Columns: query, docs, and labels
• Approximate statistics based on the first 1000 samples:

  	query	docs	labels
  type	string	list	list
  details	min: 11 characters mean: 34.2 characters max: 99 characters	min: 3 elements mean: 6.50 elements max: 10 elements	min: 3 elements mean: 6.50 elements max: 10 elements

• Samples:

  query	docs	labels
  is leukemia genetic	["1 Certain genetic disorders, such as Down syndrome, are associated with an increased risk of leukemia. 2 Exposure to certain chemicals. 3 Exposure to certain chemicals, such as benzene — which is found in gasoline and is used by the chemical industry — also is linked to an increased risk of some kinds of leukemia. 4 Smoking. 1 Previous cancer treatment. 2 People who've had certain types of chemotherapy and radiation therapy for other cancers have an increased risk of developing certain types of leukemia. 3 Genetic disorders. 4 Genetic abnormalities seem to play a role in the development", 'Causes of Leukemia. Although researchers have studied the many cellular changes associated with leukemia, it is unknown why these changes occur. It is likely that certain risk factors are involved. Many factors (e.g., age, genetics) are unmodifiable (beyond control). It is now known that all cancers, including leukemia, begin as a mutation in the genetic material—the DNA (deoxyribonucleic aci...	[1, 0, 0, 0, 0, ...]
  what does stratified seed mean	['Definition: Stratification is a means of simulating the chilling and warming that seeds would endure if left outdoors in their native climate, for the winter. Some seeds will stay dormant until triggered by a certain amount of time in cold temperature or warm, damp conditions.', 'Seeds of different species are stratified at different temperatures and for different periods of time. Some need periods of cold temperatures, between 34 and 41 degrees Fahrenheit, while some need warm stratification at 68 to 84 degrees. The ginkgo is a deciduous gymnosperm, more closely related to the pine than an oak tree. The ginkgo or maidenhair tree (Ginkgo biloba) is a living fossil. The earliest known fossil leaves of the ginkgo are 270 million years old.', 'In the wild, seed dormancy is usually overcome by the seed spending time in the ground through a winter period and having its hard seed coat softened up by frost and weathering action. By doing so the seed is undergoing a natural form of stratific...	[1, 0, 0, 0, 0, ...]
  what is a calprotectin stool test tell	['Calprotectin is a stool (fecal) test that is used to detect inflammation in the intestines. Intestinal inflammation is associated with, for example, some bacterial infections and, in people with inflammatory bowel disease (IBD) , it is associated with disease activity and severity. The test may be ordered along with other stool tests, such as a stool culture to detect a bacterial infection, a stool white blood cell test, and/or a fecal occult blood test (FOBT) .', 'Faecal calprotectin is a substance that is released into the intestines in excess when there is any inflammation there. Its presence can mean a person has an inflammatory bowel disease such as Crohn’s disease or ulcerative colitis. These conditions can cause very similar symptoms to irritable bowel syndrome. ', "Faecal calprotectin is a biochemical measurement of calprotectin in the stool. Elevated faecal calprotectin indicates the migration of neutrophils to the intestinal mucosa, which occurs during intestinal inflammati...	[1, 0, 0, 0, 0, ...]

• Loss: ListNetLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "mini_batch_size": 16
  }
  

Evaluation Dataset

ms_marco

• Dataset: ms_marco at a47ee7a
• Size: 1,000 evaluation samples
• Columns: query, docs, and labels
• Approximate statistics based on the first 1000 samples:

  	query	docs	labels
  type	string	list	list
  details	min: 9 characters mean: 33.96 characters max: 110 characters	min: 3 elements mean: 6.50 elements max: 10 elements	min: 3 elements mean: 6.50 elements max: 10 elements

• Samples:

  query	docs	labels
  what is the name of the vaccine for MMR	['Measles, mumps and rubella (MMR) vaccines. The MMR vaccine also comes in combination with chickenpox (MMRV) for 18 month old children and contains small amounts of each of the viruses at a reduced strength and a small amount of the antibiotic neomycin. MMR is given at four years of age to children who did not get their second MMR vaccine at 18 months of age. The four year old MMR dose ends in December 2015. All people born during or since 1966 should check their immunisation status to ensure they have had two doses of a measles containing vaccine.', 'The MMR vaccine protects children against all three diseases and is given at 12 months of age. A second dose using MMRV vaccine is given at 18 months of age to also protect children from chickenpox. MMR is given at four years of age to children who did not get their second MMR vaccine at 18 months of age. The four year old MMR dose ends in December 2015. All people born during or since 1966 should check their immunisation status to ensur...	[1, 0, 0, 0, 0, ...]
  what is the smallest shark in the world	['Tweet. The smallest shark in the world is the Dwarf Lanternshark; a dogfish that lives in the Caribbean Sea. They grow to a maximum of 17 centimeters and eat small fish and shrimp. The next smallest shark in the world is the Pygmy Shark that grows up to 25 centimeters long. These sharks are very closely related.', 'The dwarf lanternshark (Etmopterus perryi) is a little-known species of dogfish shark in the family Etmopteridae and possibly the smallest shark in the world, reaching a maximum known length of 21.2 cm (8.3 in). Perhaps the smallest living shark species, male dwarf lanternsharks mature at a length of 16–17.5 cm (6.3–6.9 in) and females from a length of 15.5 cm (6.1 in) with 19–20 cm (7.5–7.9 in) long pregnant individuals known.', 'The Pale Catshark at 8.27 is the smallest shark in the world. The Pale Catshark is a rare cat shark of the Scyliorhinidae family and only specimen found on the Makassar Strait. ', "The world's biggest school of sharks was observed in Galapagos, i...	[1, 0, 0, 0, 0, ...]
  how much does nicu cost per day	['According to a 2010 article in Managed Care Magazine, the average cost for infants in NICU is around $3,000 per day. This average does not include surgeries or helicopter transports.', '1 The website Ncbi.nlm.nih.gov provides information regarding the NICU; the cost according to them can reach $1,800 to $2,500 per night of care. 2 ASPE.gov is another site that provides information regarding NICU costs.', "My son was in the NICU for 11 days and the cost was about $80,000--if I remember right, that also includes the cost of the c-section and my hospital stay. He was not in an incubator, and I don't know how much more the charge for an incubator would be--probably minimal.", "MrsDeLaVara:Our LO 'cost' about $350,000 for 62 days in the hospital. She was in the high level care facility for about half of the time, which cost more. That includes my delivery too since it was at the same hospital. Her NICU doctors billed $1500 a day for the high level and about $700 for the step down nursery...	[1, 0, 0, 0, 0, ...]

• Loss: ListNetLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "mini_batch_size": 16
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• learning_rate: 2e-05
• num_train_epochs: 5
• seed: 12
• bf16: True
• 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: 16
• per_device_eval_batch_size: 16
• 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: 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: 5
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: True
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 12
• 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: 0
• 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
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss	Validation Loss
0.0002	1	2.4083	-
0.0203	100	2.0981	2.0954
0.0407	200	2.0854	2.0919
0.0610	300	2.0847	2.0912
0.0813	400	2.0849	2.0870
0.1016	500	2.0767	2.0875
0.1220	600	2.0827	2.0849
0.1423	700	2.0727	2.0850
0.1626	800	2.0715	2.0846
0.1830	900	2.074	2.0839
0.2033	1000	2.0749	2.0831
0.2236	1100	2.0714	2.0832
0.2440	1200	2.0775	2.0825
0.2643	1300	2.0691	2.0830
0.2846	1400	2.0648	2.0820
0.3049	1500	2.0746	2.0821
0.3253	1600	2.074	2.0821
0.3456	1700	2.077	2.0834
0.3659	1800	2.085	2.0812
0.3863	1900	2.073	2.0817
0.4066	2000	2.0781	2.0810
0.4269	2100	2.0805	2.0819
0.4472	2200	2.0735	2.0813
0.4676	2300	2.067	2.0816
0.4879	2400	2.0705	2.0813
0.5082	2500	2.0738	2.0813
0.5286	2600	2.0711	2.0811
0.5489	2700	2.0672	2.0808
0.5692	2800	2.0721	2.0807
0.5896	2900	2.0757	2.0805
0.6099	3000	2.0611	2.0808
0.6302	3100	2.0705	2.0800
0.6505	3200	2.0614	2.0802
0.6709	3300	2.0613	2.0806
0.6912	3400	2.0691	2.0806
0.7115	3500	2.0681	2.0802
0.7319	3600	2.0714	2.0798
0.7522	3700	2.0661	2.0800
0.7725	3800	2.0736	2.0808
0.7928	3900	2.071	2.0806
0.8132	4000	2.0749	2.0796
0.8335	4100	2.0672	2.0799
0.8538	4200	2.0699	2.0801
0.8742	4300	2.0803	2.0799
0.8945	4400	2.0716	2.0803
0.9148	4500	2.0754	2.0797
0.9351	4600	2.057	2.0798
0.9555	4700	2.06	2.0797
0.9758	4800	2.0673	2.0804
0.9961	4900	2.0687	2.0799
1.0165	5000	2.0666	2.0818
1.0368	5100	2.0648	2.0817
1.0571	5200	2.0606	2.0817
1.0775	5300	2.0606	2.0838
1.0978	5400	2.0645	2.0801
1.1181	5500	2.0628	2.0834
1.1384	5600	2.061	2.0811
1.1588	5700	2.0659	2.0807
1.1791	5800	2.0729	2.0813
1.1994	5900	2.0564	2.0808
1.2198	6000	2.0681	2.0805
1.2401	6100	2.0611	2.0800
1.2604	6200	2.0616	2.0805
1.2807	6300	2.0557	2.0807
1.3011	6400	2.0585	2.0802
1.3214	6500	2.0541	2.0807
1.3417	6600	2.0645	2.0813
1.3621	6700	2.0612	2.0804
1.3824	6800	2.0741	2.0804
1.4027	6900	2.0628	2.0802
1.4231	7000	2.0563	2.0808
1.4434	7100	2.0689	2.0804
1.4637	7200	2.0711	2.0812
1.4840	7300	2.0547	2.0802
1.5044	7400	2.0541	2.0810
1.5247	7500	2.0539	2.0811
1.5450	7600	2.0716	2.0803
1.5654	7700	2.0618	2.0797
1.5857	7800	2.0564	2.0801
1.6060	7900	2.0678	2.0801
1.6263	8000	2.0683	2.0803
1.6467	8100	2.07	2.0800
1.6670	8200	2.0678	2.0803
1.6873	8300	2.0646	2.0796
1.7077	8400	2.062	2.0802
1.7280	8500	2.0698	2.0804
1.7483	8600	2.0717	2.0809
1.7687	8700	2.0723	2.0803
1.7890	8800	2.0603	2.0809
1.8093	8900	2.0597	2.0808
1.8296	9000	2.0596	2.0811
1.8500	9100	2.0719	2.0812
1.8703	9200	2.0647	2.0804
1.8906	9300	2.0675	2.0802
1.9110	9400	2.0676	2.0806
1.9313	9500	2.0649	2.0808
1.9516	9600	2.0676	2.0815
1.9719	9700	2.0696	2.0803
1.9923	9800	2.0679	2.0808
2.0126	9900	2.0557	2.0840
2.0329	10000	2.0491	2.0830
2.0533	10100	2.0559	2.0871
2.0736	10200	2.0549	2.0854
2.0939	10300	2.0488	2.0856
2.1143	10400	2.0371	2.0879
2.1346	10500	2.0427	2.0855
2.1549	10600	2.0397	2.0842
2.1752	10700	2.0496	2.0870
2.1956	10800	2.0496	2.0860
2.2159	10900	2.0485	2.0863
2.2362	11000	2.0577	2.0845
2.2566	11100	2.0532	2.0851
2.2769	11200	2.052	2.0876
2.2972	11300	2.05	2.0873
2.3175	11400	2.0505	2.0845
2.3379	11500	2.0535	2.0860
2.3582	11600	2.0524	2.0861
2.3785	11700	2.0532	2.0858
2.3989	11800	2.0566	2.0856
2.4192	11900	2.0445	2.0862
2.4395	12000	2.0461	2.0851
2.4598	12100	2.0471	2.0852
2.4802	12200	2.0421	2.0867
2.5005	12300	2.0424	2.0854
2.5208	12400	2.0469	2.0853
2.5412	12500	2.0449	2.0855
2.5615	12600	2.0504	2.0873
2.5818	12700	2.0548	2.0856
2.6022	12800	2.0529	2.0850
2.6225	12900	2.0454	2.0858
2.6428	13000	2.0468	2.0856
2.6631	13100	2.0543	2.0869
2.6835	13200	2.0528	2.0855
2.7038	13300	2.0503	2.0863
2.7241	13400	2.0527	2.0867
2.7445	13500	2.0481	2.0852
2.7648	13600	2.0435	2.0877
2.7851	13700	2.0502	2.0875
2.8054	13800	2.052	2.0904
2.8258	13900	2.0514	2.0868
2.8461	14000	2.0476	2.0862
2.8664	14100	2.0527	2.0869
2.8868	14200	2.0503	2.0867
2.9071	14300	2.0441	2.0859
2.9274	14400	2.0474	2.0863
2.9478	14500	2.0447	2.0857
2.9681	14600	2.0515	2.0893
2.9884	14700	2.0463	2.0873
3.0087	14800	2.0458	2.0881
3.0291	14900	2.0301	2.0907
3.0494	15000	2.0343	2.0907
3.0697	15100	2.0265	2.0907
3.0901	15200	2.0338	2.0941
3.1104	15300	2.037	2.0943
3.1307	15400	2.0355	2.0954
3.1510	15500	2.0295	2.0935
3.1714	15600	2.0244	2.0936
3.1917	15700	2.0288	2.0925
3.2120	15800	2.0371	2.0915
3.2324	15900	2.0247	2.0934
3.2527	16000	2.0224	2.0929
3.2730	16100	2.0211	2.0920
3.2934	16200	2.0347	2.0936
3.3137	16300	2.0301	2.0941
3.3340	16400	2.0359	2.0955
3.3543	16500	2.0327	2.0935
3.3747	16600	2.0277	2.0935
3.3950	16700	2.0329	2.0931
3.4153	16800	2.0309	2.0958
3.4357	16900	2.0217	2.0940
3.4560	17000	2.0293	2.0942
3.4763	17100	2.0398	2.0927
3.4966	17200	2.0385	2.0951
3.5170	17300	2.0318	2.0924
3.5373	17400	2.0425	2.0968
3.5576	17500	2.0277	2.0936
3.5780	17600	2.0273	2.0942
3.5983	17700	2.0405	2.0926
3.6186	17800	2.03	2.0957
3.6390	17900	2.0198	2.0939
3.6593	18000	2.0397	2.0954
3.6796	18100	2.0378	2.0927
3.6999	18200	2.0262	2.0926
3.7203	18300	2.0321	2.0971
3.7406	18400	2.0203	2.0933
3.7609	18500	2.0473	2.0932
3.7813	18600	2.0375	2.0928
3.8016	18700	2.03	2.0955
3.8219	18800	2.0294	2.0939
3.8422	18900	2.0264	2.0930
3.8626	19000	2.0245	2.0929
3.8829	19100	2.0385	2.0925
3.9032	19200	2.0276	2.0936
3.9236	19300	2.0421	2.0928
3.9439	19400	2.0274	2.0923
3.9642	19500	2.0303	2.0940
3.9845	19600	2.0276	2.0947
4.0049	19700	2.0284	2.0941
4.0252	19800	2.0202	2.0969
4.0455	19900	2.0191	2.0955
4.0659	20000	2.0192	2.0992
4.0862	20100	2.0175	2.0964
4.1065	20200	2.021	2.0966
4.1269	20300	2.02	2.0980
4.1472	20400	2.0118	2.0983
4.1675	20500	2.02	2.0980
4.1878	20600	2.0239	2.0968
4.2082	20700	2.0148	2.0978
4.2285	20800	2.0296	2.0960
4.2488	20900	2.0078	2.0981
4.2692	21000	2.0182	2.0972
4.2895	21100	2.0184	2.0979
4.3098	21200	2.0242	2.0978
4.3301	21300	2.023	2.0969
4.3505	21400	2.0218	2.0977
4.3708	21500	2.0199	2.0984
4.3911	21600	2.0193	2.0979
4.4115	21700	2.0195	2.0983
4.4318	21800	2.0107	2.0990
4.4521	21900	2.0194	2.0973
4.4725	22000	2.022	2.0981
4.4928	22100	2.017	2.0979
4.5131	22200	2.0265	2.0991
4.5334	22300	2.0188	2.0988
4.5538	22400	2.0312	2.0979
4.5741	22500	2.0177	2.0978
4.5944	22600	2.0284	2.0982
4.6148	22700	2.0179	2.0975
4.6351	22800	2.0227	2.0960
4.6554	22900	2.0258	2.0974
4.6757	23000	2.0226	2.0974
4.6961	23100	2.0219	2.0970
4.7164	23200	2.0156	2.0974
4.7367	23300	2.0221	2.0981
4.7571	23400	2.0238	2.0982
4.7774	23500	2.0148	2.0968
4.7977	23600	2.0225	2.0983
4.8181	23700	2.0196	2.0986
4.8384	23800	2.0192	2.0987
4.8587	23900	2.0278	2.0978
4.8790	24000	2.0177	2.0982
4.8994	24100	2.0139	2.0976
4.9197	24200	2.0265	2.0984
4.9400	24300	2.0124	2.0980
4.9604	24400	2.0119	2.0978
4.9807	24500	2.0204	2.0980

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.11.13
• Sentence Transformers: 5.0.0
• Transformers: 4.51.0
• PyTorch: 2.9.1+cu126
• Accelerate: 1.8.1
• Datasets: 3.6.0
• Tokenizers: 0.21.4-dev.0

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",
}

ListNetLoss

@inproceedings{cao2007learning,
    title={Learning to Rank: From Pairwise Approach to Listwise Approach},
    author={Cao, Zhe and Qin, Tao and Liu, Tie-Yan and Tsai, Ming-Feng and Li, Hang},
    booktitle={Proceedings of the 24th international conference on Machine learning},
    pages={129--136},
    year={2007}
}