tomaarsen/csr-mxbai-embed-large-v1-gooaq-2e-4

Sparse CSR model trained on Natural Questions

This is a CSR Sparse Encoder model finetuned from mixedbread-ai/mxbai-embed-large-v1 on the gooaq dataset using the sentence-transformers library. It maps sentences & paragraphs to a 4096-dimensional sparse vector space and can be used for semantic search and sparse retrieval.

Model Details

Model Description

• Model Type: CSR Sparse Encoder
• Base model: mixedbread-ai/mxbai-embed-large-v1
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 4096 dimensions
• Similarity Function: Dot Product
• Training Dataset:
  • gooaq
• Language: en
• License: apache-2.0

Model Sources

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

Full Model Architecture

SparseEncoder(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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): CSRSparsity({'input_dim': 1024, 'hidden_dim': 4096, 'k': 256, 'k_aux': 512, 'normalize': False, 'dead_threshold': 30})
)

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 SparseEncoder

# Download from the 🤗 Hub
model = SparseEncoder("tomaarsen/csr-mxbai-embed-large-v1-gooaq-2e-4")
# Run inference
sentences = [
    'are you human korean novela?',
    "Are You Human? (Korean: 너도 인간이니; RR: Neodo Inganini; lit. Are You Human Too?) is a 2018 South Korean television series starring Seo Kang-jun and Gong Seung-yeon. It aired on KBS2's Mondays and Tuesdays at 22:00 (KST) time slot, from June 4 to August 7, 2018.",
    'A relative of European pear varieties like Bartlett and Anjou, the Asian pear is great used in recipes or simply eaten out of hand. It retains a crispness that works well in slaws and salads, and it holds its shape better than European pears when baked and cooked.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# (3, 4096)

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

Evaluation

Metrics

Sparse Information Retrieval

• Datasets: NanoMSMARCO_128, NanoNFCorpus_128 and NanoNQ_128
• Evaluated with SparseInformationRetrievalEvaluator with these parameters:

  {
      "max_active_dims": 128
  }
  

Metric	NanoMSMARCO_128	NanoNFCorpus_128	NanoNQ_128
dot_accuracy@1	0.42	0.28	0.46
dot_accuracy@3	0.64	0.46	0.62
dot_accuracy@5	0.68	0.58	0.7
dot_accuracy@10	0.8	0.66	0.82
dot_precision@1	0.42	0.28	0.46
dot_precision@3	0.2133	0.2867	0.2067
dot_precision@5	0.136	0.28	0.14
dot_precision@10	0.08	0.246	0.082
dot_recall@1	0.42	0.0101	0.44
dot_recall@3	0.64	0.0497	0.58
dot_recall@5	0.68	0.0768	0.65
dot_recall@10	0.8	0.1079	0.76
dot_ndcg@10	0.6079	0.2711	0.5977
dot_mrr@10	0.5469	0.3952	0.5692
dot_map@100	0.5547	0.1088	0.5513
row_non_zero_mean_query	128.0	128.0	128.0
row_sparsity_mean_query	0.9688	0.9688	0.9688
row_non_zero_mean_corpus	128.0	128.0	128.0
row_sparsity_mean_corpus	0.9688	0.9688	0.9688

Sparse Nano BEIR

• Dataset: NanoBEIR_mean_128
• Evaluated with SparseNanoBEIREvaluator with these parameters:

  {
      "dataset_names": [
          "msmarco",
          "nfcorpus",
          "nq"
      ],
      "max_active_dims": 128
  }
  

Metric	Value
dot_accuracy@1	0.3867
dot_accuracy@3	0.5733
dot_accuracy@5	0.6533
dot_accuracy@10	0.76
dot_precision@1	0.3867
dot_precision@3	0.2356
dot_precision@5	0.1853
dot_precision@10	0.136
dot_recall@1	0.29
dot_recall@3	0.4232
dot_recall@5	0.4689
dot_recall@10	0.556
dot_ndcg@10	0.4922
dot_mrr@10	0.5038
dot_map@100	0.405
row_non_zero_mean_query	128.0
row_sparsity_mean_query	0.9688
row_non_zero_mean_corpus	128.0
row_sparsity_mean_corpus	0.9688

Sparse Information Retrieval

• Datasets: NanoMSMARCO_256, NanoNFCorpus_256 and NanoNQ_256
• Evaluated with SparseInformationRetrievalEvaluator with these parameters:

  {
      "max_active_dims": 256
  }
  

Metric	NanoMSMARCO_256	NanoNFCorpus_256	NanoNQ_256
dot_accuracy@1	0.42	0.32	0.42
dot_accuracy@3	0.7	0.56	0.64
dot_accuracy@5	0.76	0.62	0.68
dot_accuracy@10	0.84	0.7	0.84
dot_precision@1	0.42	0.32	0.42
dot_precision@3	0.2333	0.32	0.22
dot_precision@5	0.152	0.316	0.14
dot_precision@10	0.084	0.262	0.088
dot_recall@1	0.42	0.0304	0.4
dot_recall@3	0.7	0.0717	0.6
dot_recall@5	0.76	0.0931	0.63
dot_recall@10	0.84	0.1333	0.79
dot_ndcg@10	0.6326	0.3071	0.5943
dot_mrr@10	0.5661	0.4525	0.5506
dot_map@100	0.5727	0.143	0.533
row_non_zero_mean_query	256.0	256.0	256.0
row_sparsity_mean_query	0.9375	0.9375	0.9375
row_non_zero_mean_corpus	256.0	256.0	256.0
row_sparsity_mean_corpus	0.9375	0.9375	0.9375

Sparse Nano BEIR

• Dataset: NanoBEIR_mean_256
• Evaluated with SparseNanoBEIREvaluator with these parameters:

  {
      "dataset_names": [
          "msmarco",
          "nfcorpus",
          "nq"
      ],
      "max_active_dims": 256
  }
  

Metric	Value
dot_accuracy@1	0.3867
dot_accuracy@3	0.6333
dot_accuracy@5	0.6867
dot_accuracy@10	0.7933
dot_precision@1	0.3867
dot_precision@3	0.2578
dot_precision@5	0.2027
dot_precision@10	0.1447
dot_recall@1	0.2835
dot_recall@3	0.4572
dot_recall@5	0.4944
dot_recall@10	0.5878
dot_ndcg@10	0.5113
dot_mrr@10	0.5231
dot_map@100	0.4163
row_non_zero_mean_query	256.0
row_sparsity_mean_query	0.9375
row_non_zero_mean_corpus	256.0
row_sparsity_mean_corpus	0.9375

Training Details

Training Dataset

gooaq

• Dataset: gooaq at b089f72
• Size: 3,011,496 training samples
• Columns: question and answer
• Approximate statistics based on the first 1000 samples:

  	question	answer
  type	string	string
  details	min: 8 tokens mean: 11.87 tokens max: 23 tokens	min: 14 tokens mean: 60.09 tokens max: 201 tokens

• Samples:

  question	answer
  what is the difference between clay and mud mask?	The main difference between the two is that mud is a skin-healing agent, while clay is a cosmetic, drying agent. Clay masks are most useful for someone who has oily skin and is prone to breakouts of acne and blemishes.
  myki how much on card?	A full fare myki card costs $6 and a concession, seniors or child myki costs $3. For more information about how to use your myki, visit ptv.vic.gov.au or call 1800 800 007.
  how to find out if someone blocked your phone number on iphone?	If you get a notification like "Message Not Delivered" or you get no notification at all, that's a sign of a potential block. Next, you could try calling the person. If the call goes right to voicemail or rings once (or a half ring) then goes to voicemail, that's further evidence you may have been blocked.

• Loss: CSRLoss with these parameters:

  {
      "beta": 0.1,
      "gamma": 1.0,
      "loss": "SparseMultipleNegativesRankingLoss(scale=1.0, similarity_fct='dot_score')"
  }
  

Evaluation Dataset

gooaq

• Dataset: gooaq at b089f72
• Size: 1,000 evaluation samples
• Columns: question and answer
• Approximate statistics based on the first 1000 samples:

  	question	answer
  type	string	string
  details	min: 8 tokens mean: 11.88 tokens max: 22 tokens	min: 14 tokens mean: 61.03 tokens max: 127 tokens

• Samples:

  question	answer
  how do i program my directv remote with my tv?	['Press MENU on your remote.', 'Select Settings & Help > Settings > Remote Control > Program Remote.', 'Choose the device (TV, audio, DVD) you wish to program. ... ', 'Follow the on-screen prompts to complete programming.']
  are rodrigues fruit bats nocturnal?	Before its numbers were threatened by habitat destruction, storms, and hunting, some of those groups could number 500 or more members. Sunrise, sunset. Rodrigues fruit bats are most active at dawn, at dusk, and at night.
  why does your heart rate increase during exercise bbc bitesize?	During exercise there is an increase in physical activity and muscle cells respire more than they do when the body is at rest. The heart rate increases during exercise. The rate and depth of breathing increases - this makes sure that more oxygen is absorbed into the blood, and more carbon dioxide is removed from it.

• Loss: CSRLoss with these parameters:

  {
      "beta": 0.1,
      "gamma": 1.0,
      "loss": "SparseMultipleNegativesRankingLoss(scale=1.0, similarity_fct='dot_score')"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• learning_rate: 0.0002
• num_train_epochs: 1
• warmup_ratio: 0.1
• bf16: True
• load_best_model_at_end: True
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• 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: 0.0002
• 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: 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}
• 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
• dispatch_batches: None
• split_batches: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss	Validation Loss	NanoMSMARCO_128_dot_ndcg@10	NanoNFCorpus_128_dot_ndcg@10	NanoNQ_128_dot_ndcg@10	NanoBEIR_mean_128_dot_ndcg@10	NanoMSMARCO_256_dot_ndcg@10	NanoNFCorpus_256_dot_ndcg@10	NanoNQ_256_dot_ndcg@10	NanoBEIR_mean_256_dot_ndcg@10
-1	-1	-	-	0.6175	0.2875	0.5432	0.4827	0.6158	0.3234	0.5929	0.5107
0.0064	300	0.3621	-	-	-	-	-	-	-	-	-
0.0128	600	0.3319	-	-	-	-	-	-	-	-	-
0.0191	900	0.3212	-	-	-	-	-	-	-	-	-
0.0255	1200	0.3154	-	-	-	-	-	-	-	-	-
0.0319	1500	0.3129	-	-	-	-	-	-	-	-	-
0.0383	1800	0.309	-	-	-	-	-	-	-	-	-
0.0446	2100	0.317	-	-	-	-	-	-	-	-	-
0.0510	2400	0.2997	-	-	-	-	-	-	-	-	-
0.0574	2700	0.3409	-	-	-	-	-	-	-	-	-
0.0638	3000	0.3251	0.3136	0.6049	0.2393	0.5583	0.4675	0.5950	0.2559	0.5555	0.4688
0.0701	3300	0.3291	-	-	-	-	-	-	-	-	-
0.0765	3600	0.3366	-	-	-	-	-	-	-	-	-
0.0829	3900	0.3286	-	-	-	-	-	-	-	-	-
0.0893	4200	0.3264	-	-	-	-	-	-	-	-	-
0.0956	4500	0.3413	-	-	-	-	-	-	-	-	-
0.1020	4800	0.3352	-	-	-	-	-	-	-	-	-
0.1084	5100	0.3323	-	-	-	-	-	-	-	-	-
0.1148	5400	0.3308	-	-	-	-	-	-	-	-	-
0.1211	5700	0.3127	-	-	-	-	-	-	-	-	-
0.1275	6000	0.3224	0.2949	0.5445	0.2155	0.5394	0.4331	0.5911	0.2340	0.5365	0.4539
0.1339	6300	0.3216	-	-	-	-	-	-	-	-	-
0.1403	6600	0.3202	-	-	-	-	-	-	-	-	-
0.1466	6900	0.3296	-	-	-	-	-	-	-	-	-
0.1530	7200	0.3171	-	-	-	-	-	-	-	-	-
0.1594	7500	0.3141	-	-	-	-	-	-	-	-	-
0.1658	7800	0.3202	-	-	-	-	-	-	-	-	-
0.1721	8100	0.3088	-	-	-	-	-	-	-	-	-
0.1785	8400	0.304	-	-	-	-	-	-	-	-	-
0.1849	8700	0.3105	-	-	-	-	-	-	-	-	-
0.1913	9000	0.307	0.2849	0.6038	0.2258	0.5471	0.4589	0.6241	0.2449	0.5498	0.4730
0.1976	9300	0.3043	-	-	-	-	-	-	-	-	-
0.2040	9600	0.3035	-	-	-	-	-	-	-	-	-
0.2104	9900	0.3069	-	-	-	-	-	-	-	-	-
0.2168	10200	0.3174	-	-	-	-	-	-	-	-	-
0.2231	10500	0.3111	-	-	-	-	-	-	-	-	-
0.2295	10800	0.295	-	-	-	-	-	-	-	-	-
0.2359	11100	0.2892	-	-	-	-	-	-	-	-	-
0.2423	11400	0.3012	-	-	-	-	-	-	-	-	-
0.2486	11700	0.3061	-	-	-	-	-	-	-	-	-
0.2550	12000	0.2863	0.2631	0.6190	0.2720	0.5379	0.4763	0.6056	0.2898	0.5419	0.4791
0.2614	12300	0.3008	-	-	-	-	-	-	-	-	-
0.2678	12600	0.2849	-	-	-	-	-	-	-	-	-
0.2741	12900	0.2876	-	-	-	-	-	-	-	-	-
0.2805	13200	0.2963	-	-	-	-	-	-	-	-	-
0.2869	13500	0.2926	-	-	-	-	-	-	-	-	-
0.2933	13800	0.2855	-	-	-	-	-	-	-	-	-
0.2996	14100	0.2868	-	-	-	-	-	-	-	-	-
0.3060	14400	0.294	-	-	-	-	-	-	-	-	-
0.3124	14700	0.3008	-	-	-	-	-	-	-	-	-
0.3188	15000	0.293	0.2745	0.5538	0.2847	0.5422	0.4602	0.5615	0.2976	0.5588	0.4726
0.3252	15300	0.2776	-	-	-	-	-	-	-	-	-
0.3315	15600	0.2906	-	-	-	-	-	-	-	-	-
0.3379	15900	0.2874	-	-	-	-	-	-	-	-	-
0.3443	16200	0.2834	-	-	-	-	-	-	-	-	-
0.3507	16500	0.2718	-	-	-	-	-	-	-	-	-
0.3570	16800	0.2834	-	-	-	-	-	-	-	-	-
0.3634	17100	0.2833	-	-	-	-	-	-	-	-	-
0.3698	17400	0.281	-	-	-	-	-	-	-	-	-
0.3762	17700	0.2922	-	-	-	-	-	-	-	-	-
0.3825	18000	0.279	0.2623	0.5851	0.2696	0.5097	0.4548	0.5849	0.2776	0.5570	0.4732
0.3889	18300	0.2894	-	-	-	-	-	-	-	-	-
0.3953	18600	0.283	-	-	-	-	-	-	-	-	-
0.4017	18900	0.2824	-	-	-	-	-	-	-	-	-
0.4080	19200	0.2758	-	-	-	-	-	-	-	-	-
0.4144	19500	0.2893	-	-	-	-	-	-	-	-	-
0.4208	19800	0.278	-	-	-	-	-	-	-	-	-
0.4272	20100	0.2814	-	-	-	-	-	-	-	-	-
0.4335	20400	0.278	-	-	-	-	-	-	-	-	-
0.4399	20700	0.2783	-	-	-	-	-	-	-	-	-
0.4463	21000	0.2803	0.2510	0.5880	0.2664	0.5664	0.4736	0.6115	0.2734	0.5465	0.4772
0.4527	21300	0.2668	-	-	-	-	-	-	-	-	-
0.4590	21600	0.2828	-	-	-	-	-	-	-	-	-
0.4654	21900	0.2815	-	-	-	-	-	-	-	-	-
0.4718	22200	0.2778	-	-	-	-	-	-	-	-	-
0.4782	22500	0.271	-	-	-	-	-	-	-	-	-
0.4845	22800	0.2696	-	-	-	-	-	-	-	-	-
0.4909	23100	0.2698	-	-	-	-	-	-	-	-	-
0.4973	23400	0.2768	-	-	-	-	-	-	-	-	-
0.5037	23700	0.2626	-	-	-	-	-	-	-	-	-
0.5100	24000	0.2611	0.2414	0.6078	0.2635	0.5668	0.4794	0.6231	0.2942	0.5944	0.5039
0.5164	24300	0.2736	-	-	-	-	-	-	-	-	-
0.5228	24600	0.2695	-	-	-	-	-	-	-	-	-
0.5292	24900	0.2673	-	-	-	-	-	-	-	-	-
0.5355	25200	0.2746	-	-	-	-	-	-	-	-	-
0.5419	25500	0.2681	-	-	-	-	-	-	-	-	-
0.5483	25800	0.2676	-	-	-	-	-	-	-	-	-
0.5547	26100	0.2686	-	-	-	-	-	-	-	-	-
0.5610	26400	0.2652	-	-	-	-	-	-	-	-	-
0.5674	26700	0.2596	-	-	-	-	-	-	-	-	-
0.5738	27000	0.2677	0.2494	0.6018	0.2460	0.5280	0.4586	0.6238	0.2775	0.5673	0.4895
0.5802	27300	0.2621	-	-	-	-	-	-	-	-	-
0.5865	27600	0.2558	-	-	-	-	-	-	-	-	-
0.5929	27900	0.251	-	-	-	-	-	-	-	-	-
0.5993	28200	0.2601	-	-	-	-	-	-	-	-	-
0.6057	28500	0.2612	-	-	-	-	-	-	-	-	-
0.6120	28800	0.2695	-	-	-	-	-	-	-	-	-
0.6184	29100	0.2662	-	-	-	-	-	-	-	-	-
0.6248	29400	0.2589	-	-	-	-	-	-	-	-	-
0.6312	29700	0.2602	-	-	-	-	-	-	-	-	-
0.6376	30000	0.2698	0.2507	0.5892	0.2996	0.5386	0.4758	0.6102	0.2941	0.5535	0.4860
0.6439	30300	0.2625	-	-	-	-	-	-	-	-	-
0.6503	30600	0.2598	-	-	-	-	-	-	-	-	-
0.6567	30900	0.2594	-	-	-	-	-	-	-	-	-
0.6631	31200	0.2618	-	-	-	-	-	-	-	-	-
0.6694	31500	0.2556	-	-	-	-	-	-	-	-	-
0.6758	31800	0.2591	-	-	-	-	-	-	-	-	-
0.6822	32100	0.2544	-	-	-	-	-	-	-	-	-
0.6886	32400	0.2589	-	-	-	-	-	-	-	-	-
0.6949	32700	0.2522	-	-	-	-	-	-	-	-	-
0.7013	33000	0.2521	0.2535	0.6053	0.2650	0.5329	0.4677	0.6115	0.2925	0.6057	0.5032
0.7077	33300	0.2576	-	-	-	-	-	-	-	-	-
0.7141	33600	0.2582	-	-	-	-	-	-	-	-	-
0.7204	33900	0.2567	-	-	-	-	-	-	-	-	-
0.7268	34200	0.2577	-	-	-	-	-	-	-	-	-
0.7332	34500	0.2568	-	-	-	-	-	-	-	-	-
0.7396	34800	0.254	-	-	-	-	-	-	-	-	-
0.7459	35100	0.2489	-	-	-	-	-	-	-	-	-
0.7523	35400	0.2545	-	-	-	-	-	-	-	-	-
0.7587	35700	0.2476	-	-	-	-	-	-	-	-	-
0.7651	36000	0.2637	0.2397	0.6138	0.2726	0.5627	0.4831	0.6056	0.2889	0.5745	0.4897
0.7714	36300	0.2508	-	-	-	-	-	-	-	-	-
0.7778	36600	0.2569	-	-	-	-	-	-	-	-	-
0.7842	36900	0.2419	-	-	-	-	-	-	-	-	-
0.7906	37200	0.2453	-	-	-	-	-	-	-	-	-
0.7969	37500	0.2456	-	-	-	-	-	-	-	-	-
0.8033	37800	0.2497	-	-	-	-	-	-	-	-	-
0.8097	38100	0.2556	-	-	-	-	-	-	-	-	-
0.8161	38400	0.252	-	-	-	-	-	-	-	-	-
0.8224	38700	0.2423	-	-	-	-	-	-	-	-	-
0.8288	39000	0.2545	0.2301	0.5927	0.2895	0.5553	0.4792	0.5979	0.2987	0.5587	0.4851
0.8352	39300	0.2482	-	-	-	-	-	-	-	-	-
0.8416	39600	0.2429	-	-	-	-	-	-	-	-	-
0.8479	39900	0.2463	-	-	-	-	-	-	-	-	-
0.8543	40200	0.2354	-	-	-	-	-	-	-	-	-
0.8607	40500	0.2466	-	-	-	-	-	-	-	-	-
0.8671	40800	0.2484	-	-	-	-	-	-	-	-	-
0.8734	41100	0.2448	-	-	-	-	-	-	-	-	-
0.8798	41400	0.2448	-	-	-	-	-	-	-	-	-
0.8862	41700	0.2515	-	-	-	-	-	-	-	-	-
0.8926	42000	0.2428	0.2392	0.6001	0.2826	0.5857	0.4895	0.6208	0.3019	0.6010	0.5079
0.8989	42300	0.2497	-	-	-	-	-	-	-	-	-
0.9053	42600	0.2415	-	-	-	-	-	-	-	-	-
0.9117	42900	0.2408	-	-	-	-	-	-	-	-	-
0.9181	43200	0.242	-	-	-	-	-	-	-	-	-
0.9245	43500	0.2412	-	-	-	-	-	-	-	-	-
0.9308	43800	0.2472	-	-	-	-	-	-	-	-	-
0.9372	44100	0.2408	-	-	-	-	-	-	-	-	-
0.9436	44400	0.2374	-	-	-	-	-	-	-	-	-
0.9500	44700	0.2312	-	-	-	-	-	-	-	-	-
0.9563	45000	0.2412	0.2379	0.6079	0.2711	0.5977	0.4922	0.6326	0.3071	0.5943	0.5113
0.9627	45300	0.2381	-	-	-	-	-	-	-	-	-
0.9691	45600	0.2456	-	-	-	-	-	-	-	-	-
0.9755	45900	0.2418	-	-	-	-	-	-	-	-	-
0.9818	46200	0.2355	-	-	-	-	-	-	-	-	-
0.9882	46500	0.2424	-	-	-	-	-	-	-	-	-
0.9946	46800	0.2389	-	-	-	-	-	-	-	-	-

• The bold row denotes the saved checkpoint.

Environmental Impact

Carbon emissions were measured using CodeCarbon.

• Energy Consumed: 1.202 kWh
• Carbon Emitted: 0.467 kg of CO2
• Hours Used: 3.125 hours

Training Hardware

• On Cloud: No
• GPU Model: 1 x NVIDIA GeForce RTX 3090
• CPU Model: 13th Gen Intel(R) Core(TM) i7-13700K
• RAM Size: 31.78 GB

Framework Versions

• Python: 3.11.6
• Sentence Transformers: 4.2.0.dev0
• Transformers: 4.49.0
• PyTorch: 2.6.0+cu124
• Accelerate: 1.5.1
• Datasets: 2.21.0
• 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",
}

CSRLoss

@misc{wen2025matryoshkarevisitingsparsecoding,
      title={Beyond Matryoshka: Revisiting Sparse Coding for Adaptive Representation},
      author={Tiansheng Wen and Yifei Wang and Zequn Zeng and Zhong Peng and Yudi Su and Xinyang Liu and Bo Chen and Hongwei Liu and Stefanie Jegelka and Chenyu You},
      year={2025},
      eprint={2503.01776},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2503.01776},
}

SparseMultipleNegativesRankingLoss

@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}
}