Upload rapha-embed-clinical-v1 (fine-tuned + GGUF)

4927cad verified about 1 month ago

20.6 kB

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:132037
  - loss:MultipleNegativesRankingLoss
base_model: BAAI/bge-small-en-v1.5
widget:
  - source_sentence: back pain. I'm not sure what to make of it.
    sentences:
      - 'Observed: back pain — musculoskeletal system'
      - Patient is frustrated with the medical system
      - 'OPQRST — Severity: fever rated 7/10 by patient'
  - source_sentence: >-
      I'm absolutely terrified. I've had eye redness for a few days and I keep
      imagining the worst.
    sentences:
      - >-
        Observed: palpitations and irregular heart rate — cardiovascular —
        physician confirmed: Atrial fibrillation
      - 'Patient presented: blood in urine — neurological/cognitive system'
      - >-
        Chief complaint: patient expressed concern, reason not yet specified —
        calm presentation
  - source_sentence: I take a pill called clopidogrel — not sure of the dose
    sentences:
      - Patient is drug-free by choice — lifestyle management only
      - >-
        Observed: fatigue and shortness of breath — haematological — physician
        confirmed: Anaemia
      - 'Current medication: clopidogrel 75mg daily — antiplatelet — for post-MI'
  - source_sentence: I'm on gabapentin for my neuropathic pain
    sentences:
      - Patient takes no regular medications
      - >-
        Current medication: gabapentin 300mg three times daily — anticonvulsant
        — for neuropathic pain
      - 'Observed: palpitations — cardiovascular system'
  - source_sentence: my breathing is fast and I feel disoriented
    sentences:
      - Patient is anxious and hyperventilating due to health anxiety
      - 'OPQRST — Onset: abdominal pain started suddenly without warning'
      - >-
        Red flag: tachypnoea + confusion — qSOFA >= 2 — possible sepsis —
        IMMEDIATE ESCALATION
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy
model-index:
  - name: SentenceTransformer based on BAAI/bge-small-en-v1.5
    results:
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: rapha val
          type: rapha-val
        metrics:
          - type: cosine_accuracy
            value: 1
            name: Cosine Accuracy

SentenceTransformer based on BAAI/bge-small-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-small-en-v1.5. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for retrieval.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: BAAI/bge-small-en-v1.5
Maximum Sequence Length: 128 tokens
Output Dimensionality: 384 dimensions
Similarity Function: Cosine Similarity
Supported Modality: Text

Model Sources

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

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'BertModel'})
  (1): Pooling({'embedding_dimension': 384, 'pooling_mode': 'cls', 'include_prompt': True})
  (2): Normalize({})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'my breathing is fast and I feel disoriented',
    'Red flag: tachypnoea + confusion — qSOFA >= 2 — possible sepsis — IMMEDIATE ESCALATION',
    'Patient is anxious and hyperventilating due to health anxiety',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.8755, 0.1384],
#         [0.8755, 1.0000, 0.0772],
#         [0.1384, 0.0772, 1.0000]])

Evaluation

Metrics

Triplet

Dataset: rapha-val
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	1.0

Training Details

Training Dataset

Unnamed Dataset

Size: 132,037 training samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 100 samples:

	anchor	positive	negative
type	string	string	string
modality	text	text	text
details	min: 5 tokens mean: 16.2 tokens max: 34 tokens	min: 8 tokens mean: 20.9 tokens max: 33 tokens	min: 8 tokens mean: 12.62 tokens max: 19 tokens

Samples:

anchor	positive	negative
`I'm trying to stay calm but I have tremors and I genuinely feel like something is very wrong.`	`Observed: tremors — neurological system`	`Patient is defensive, resistant to clinical interview`
`neck pain so bad I can't look down and light sensitivity too`	`Red flag: neck stiffness + photophobia + fever — possible meningitis — IMMEDIATE ESCALATION`	`Patient reports muscle tension in neck from bad posture`
`neck pain so bad I can't look down and light sensitivity too`	`Red flag: neck stiffness + photophobia + fever — possible meningitis — IMMEDIATE ESCALATION`	`Patient reports muscle tension in neck from bad posture`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "gather_across_devices": false,
    "directions": [
        "query_to_doc"
    ],
    "partition_mode": "joint",
    "hardness_mode": null,
    "hardness_strength": 0.0
}

Evaluation Dataset

Unnamed Dataset

Size: 15,000 evaluation samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 100 samples:

	anchor	positive	negative
type	string	string	string
modality	text	text	text
details	min: 5 tokens mean: 15.39 tokens max: 27 tokens	min: 9 tokens mean: 18.62 tokens max: 33 tokens	min: 7 tokens mean: 12.74 tokens max: 23 tokens

Samples:

anchor	positive	negative
`I'm on prednisolone 5mg daily`	`Current medication: prednisolone 5mg daily — corticosteroid — for inflammatory condition`	`Patient is not currently prescribed anything`
`Diagnosis recorded: Community-acquired pneumonia — ICD-10 code J18.9`	`Observed: cough, fever, shortness of breath — respiratory/infectious — physician confirmed: Community-acquired pneumonia`	`Observed: cough — upper respiratory tract infection — self-limiting, no antibiotics`
`breast lump.`	`Session start: patient presented with unspecified concern — Minimal communication style`	`Patient presented: numbness in limbs — respiratory system`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "gather_across_devices": false,
    "directions": [
        "query_to_doc"
    ],
    "partition_mode": "joint",
    "hardness_mode": null,
    "hardness_strength": 0.0
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 256
learning_rate: 2e-05
lr_scheduler_type: cosine
warmup_steps: 0.1
bf16: True
per_device_eval_batch_size: 256
load_best_model_at_end: True
dataloader_drop_last: True

All Hyperparameters

Click to expand

per_device_train_batch_size: 256
num_train_epochs: 3
max_steps: -1
learning_rate: 2e-05
lr_scheduler_type: cosine
lr_scheduler_kwargs: None
warmup_steps: 0.1
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: None
trackio_bucket_id: None
trackio_static_space_id: None
per_device_eval_batch_size: 256
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: True
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: 0
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_static_graph: None
ddp_backend: None
ddp_timeout: 1800
fsdp: None
fsdp_config: None
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

Epoch	Step	Training Loss	Validation Loss	rapha-val_cosine_accuracy
0.0971	50	3.5863	-	-
0.1942	100	2.3320	-	-
0.2913	150	1.7194	-	-
0.3883	200	1.5596	-	-
0.4854	250	1.4893	-	-
0.5825	300	1.4520	-	-
0.6796	350	1.4428	-	-
0.7767	400	1.4143	-	-
0.8738	450	1.4227	-	-
0.9709	500	1.3991	1.4893	1.0
1.0680	550	1.3811	-	-
1.1650	600	1.3818	-	-
1.2621	650	1.3876	-	-
1.3592	700	1.3830	-	-
1.4563	750	1.3690	-	-
1.5534	800	1.3927	-	-
1.6505	850	1.3683	-	-
1.7476	900	1.3772	-	-
1.8447	950	1.3812	-	-
1.9417	1000	1.4034	1.4836	1.0
2.0388	1050	1.3812	-	-
2.1359	1100	1.3562	-	-
2.2330	1150	1.3676	-	-
2.3301	1200	1.3813	-	-
2.4272	1250	1.3849	-	-
2.5243	1300	1.3729	-	-
2.6214	1350	1.3770	-	-
2.7184	1400	1.3665	-	-
2.8155	1450	1.3469	-	-
2.9126	1500	1.3790	1.4779	1.0
3.0	1545	-	1.4779	1.0

The bold row denotes the saved checkpoint.

Training Time

Training: 6.3 minutes

Framework Versions

Python: 3.12.13
Sentence Transformers: 5.6.0
Transformers: 5.12.1
PyTorch: 2.11.0+cu128
Accelerate: 1.14.0
Datasets: 4.0.0
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",
}

MultipleNegativesRankingLoss

@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}