SentenceTransformer based on distilbert/distilroberta-base

This is a sentence-transformers model finetuned from distilbert/distilroberta-base. It maps sentences & paragraphs to a 768-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: distilbert/distilroberta-base
Maximum Sequence Length: 128 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity

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': 128, 'do_lower_case': False}) with Transformer model: RobertaModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, '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("nikatonika/chatbot_sentence-transformer")
# Run inference
sentences = [
    'Like you are now? [SEP] That liver is going to somebody right now. Were doing that surgery. If you do the surgery, youll be killing a mother of four. Father of three. I was guessing. Naphthalene poisoning is the best explanation we have for whats wrong with your son. It explains the internal bleeding, the hemolytic anemia, the liver failure! it also predicts whatll happen next. If you do the surgery hes gonna lay on that table for fourteen hours while his body continues to burn fat and release poison into his system. Either way, I did you a favor. Hes awake now, youve got a chance to say goodbye.',
    'If you do the surgery hes gonna lay on that table for fourteen hours while his body continues to burn fat and release poison into',
    'I know none of that. If I did, youd be the last to know.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

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

Evaluation

Metrics

Triplet

Dataset: dev_evaluator
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9809

Training Details

Training Dataset

Unnamed Dataset

Size: 6,284 training samples
Columns: sentence_0, sentence_1, and sentence_2

Approximate statistics based on the first 1000 samples:

	sentence_0	sentence_1	sentence_2
type	string	string	string
details	min: 32 tokens mean: 106.25 tokens max: 128 tokens	min: 4 tokens mean: 20.84 tokens max: 128 tokens	min: 5 tokens mean: 18.96 tokens max: 69 tokens

Samples:

sentence_0	sentence_1	sentence_2
`I thought Everybody lied? [SEP] Told you, cant trust people. She pRobably knew she was allergic to gadolinium, figured it was an easy way to get someone to cut a hole in her throat. Cant get a picture, gonna have to get a thousand words. You actually want me to talk to the Patient? Get a history? We need to know if theres some genetic or environmental causes triggering an inflammatory response. Truth begins in lies. Think about it.`	`Truth begins in lies. Think about it.`	`the Krusshy and the... Krab... pizza...`
Whats that? [SEP] Her blood pressures rising. Mines rising too, course I am doing battle with a deity. In the heart, injecting the dye. Right coronary flow isnt obstructed, left coronary flow looks normal. Looks like youre wrong. Either Im right, or this test is about to go very bad. She has one... two... third ostium. How Many is she supposed to have? Dos. All the third ones doing is causing inflammation, throwing off clots, giving away the angiogram. No huMan would screw up that big! Dont worry, just one more surgery and youll be fine.	`She has one... two... third ostium. How Many is she supposed to have? Dos. All the third ones doing is causing inflammation, throwing off clots, giving away the angiogram.`	`Of course I’m jokin’! I don’t take checks.`
Do me a favor!? [SEP] Mmhhmmm, I need to go peepee. Dial it up a notch and repeat. Ill be back. Ooh, girl in the boys bathroom. Very dramatic. Must be very important what you have to say to me. Yesterday your Patients tumor was 5.8 centimeters. Today its 4.6. How did that happen? At a guess, Id say Dr. House must be really really good ì why am I wasting him on hiccups?ù I wash before and after. You also requisitioned 20cc of ethanol what Patient was that for? Or are you planning a party? I was gonna say leave,ù but that works.	`I was gonna say leave,ù but that works.`	`I seem to recall them giving you a bit of trouble as well.`

Loss: TripletLoss with these parameters:

{
    "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
    "triplet_margin": 5
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
num_train_epochs: 1
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 8
per_device_eval_batch_size: 8
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: 5e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1
num_train_epochs: 1
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: 42
data_seed: None
jit_mode_eval: False
use_ipex: False
bf16: False
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: False
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: batch_sampler
multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss	dev_evaluator_cosine_accuracy
-1	-1	-	0.7078
0.2545	200	-	0.9255
0.5089	400	-	0.9701
0.6361	500	1.6621	-
0.7634	600	-	0.9752
1.0	786	-	0.9790
-1	-1	-	0.9790
0.2545	200	-	0.9752
0.5089	400	-	0.9790
0.6361	500	0.298	-
0.7634	600	-	0.9790
1.0	786	-	0.9803
-1	-1	-	0.9803
0.2545	200	-	0.9777
0.5089	400	-	0.9796
0.6361	500	0.0783	-
0.7634	600	-	0.9809

Framework Versions

Python: 3.11.11
Sentence Transformers: 3.4.1
Transformers: 4.49.0
PyTorch: 2.6.0+cu124
Accelerate: 1.3.0
Datasets: 3.3.2
Tokenizers: 0.21.0

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",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification},
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}