NeurologyAI/mxbai-remedy-matryoshka_v3

MXBD REMEDY Matryoshka_v2

This is a sentence-transformers model finetuned from mixedbread-ai/mxbai-embed-large-v1 on the json dataset. It maps sentences & paragraphs to a 1024-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: mixedbread-ai/mxbai-embed-large-v1
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • json
• Language: en
• License: apache-2.0

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (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})
)

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("NeurologyAI/mxbai-remedy-matryoshka_v3")
# Run inference
sentences = [
    "[TextBlock(text='What is critical to avoid misinterpretation of immunohistochemical stainings of glioma tissue?', type='text')]",
    'To avoid misinterpretation of immunohistochemical stainings of glioma tissue it is critical to choose antibodies that work well on FFPE material, to optimize and validate tissue pretreatment and staining protocols stringently, and to perform tests alongside appropriate negative and positive controls. For example, ATRX and H3 p.K28me3 immunohistochemistry are sensitive to hypoxia and crush artifacts; in this setting nuclear staining in non-neoplastic cells can serve as a positive internal control.',
    'Cerebral MRI should include axial T1-weighted, axial FLAIR, axial diffusion-weighted, axial T2-weighted, post-gadolinium 3D T1-weighted and post-gadolinium 3D FLAIR sequences. Spinal MRI should include post-gadolinium sagittal T1-weighted sequences. Spine sagittal T1-weighted sequences without contrast and sagittal fat-suppression T2-weighted sequences, combined with axial T1-weighted images with contrast of regions of interest, may also be considered.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Evaluation

Metrics

Information Retrieval

• Datasets: dim_768, dim_512, dim_256, dim_128, dim_64, dim_768, dim_512, dim_256, dim_128 and dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	dim_768	dim_512	dim_256	dim_128	dim_64
cosine_accuracy@1	1.0	1.0	1.0	1.0	0.9524
cosine_accuracy@3	1.0	1.0	1.0	1.0	1.0
cosine_accuracy@5	1.0	1.0	1.0	1.0	1.0
cosine_accuracy@10	1.0	1.0	1.0	1.0	1.0
cosine_precision@1	1.0	1.0	1.0	1.0	0.9524
cosine_precision@3	0.3333	0.3333	0.3333	0.3333	0.3333
cosine_precision@5	0.2	0.2	0.2	0.2	0.2
cosine_precision@10	0.1	0.1	0.1	0.1	0.1
cosine_recall@1	1.0	1.0	1.0	1.0	0.9524
cosine_recall@3	1.0	1.0	1.0	1.0	1.0
cosine_recall@5	1.0	1.0	1.0	1.0	1.0
cosine_recall@10	1.0	1.0	1.0	1.0	1.0
cosine_ndcg@10	1.0	1.0	1.0	1.0	0.9824
cosine_mrr@10	1.0	1.0	1.0	1.0	0.9762
cosine_map@100	1.0	1.0	1.0	1.0	0.9762

Training Details

Training Dataset

json

• Dataset: json
• Size: 181 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 181 samples:

  	anchor	positive
  type	string	string
  details	min: 9 tokens mean: 41.06 tokens max: 77 tokens	min: 16 tokens mean: 53.83 tokens max: 239 tokens

• Samples:

  anchor	positive
  [TextBlock(text='What is the recommended treatment for patients with recurrent or atypical meningiomas?', type='text')]	Patients with recurrent or atypical meningiomas hould receive fractionated radiotherapy.
  [TextBlock(text='What is the recommended treatment for headache in patients with gliomas?', type='text')]	Pain or headache. Corticosteroids (dexamethasone) should be the mainstay of treatment for headache in patients with gliomas. Analgesics and co-analgesics could also be considered in the treatment of headache in patients with gliomas in accordance with the WHO cancer pain ladder. During care in the end-of-life phase, consideration needs to be given to the management of headache with corticosteroids; advantages of corticosteroids (alleviation of symptoms) should be weighed against side-effects (such as delirium).
  [TextBlock(text='What is the priority in the current surgical approach to gliomas?', type='text')]	The extent of resection is a prognostic factor and thus, efforts at obtaining complete resections are justified across all glioma entities. In the current surgical approach to gliomas, the prevention of new permanent neurological deficits has higher priority than the extent of resection.

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 4
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• bf16: True
• tf32: True
• load_best_model_at_end: True
• optim: adamw_torch_fused
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• eval_accumulation_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: 4
• max_steps: -1
• lr_scheduler_type: cosine
• 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: True
• 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_fused
• 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: False
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• 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
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	dim_768_cosine_ndcg@10	dim_512_cosine_ndcg@10	dim_256_cosine_ndcg@10	dim_128_cosine_ndcg@10	dim_64_cosine_ndcg@10
1.0	1	0.9235	0.8645	0.8624	0.7593	0.6138
2.0	2	1.0	0.9649	1.0	0.9824	0.9473
3.0	4	1.0	1.0	1.0	1.0	0.9473
1.0	1	1.0	1.0	1.0	1.0	0.9473
2.0	2	1.0	1.0	1.0	1.0	0.9824
3.0	4	1.0	1.0	1.0	1.0	0.9824

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.3.1
• Transformers: 4.41.2
• PyTorch: 2.1.2+cu121
• Accelerate: 1.1.1
• Datasets: 2.19.1
• Tokenizers: 0.19.1

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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning},
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

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}
}