SentenceTransformer

This model aims at encoding text information from deviations Titles and/or Deviation Description (Event) for various Takeda site.

This is a sentence-transformers model trained. 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
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 tokens
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': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (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("sentence_transformers_model_id")
# Run inference
sentences = [
    'Emergency Door in Staircase Room 1044 for Post-Viral Found Not Completely Closed',
    'On 02Jul2023, Manufacturing Supervisor (EID 50251544) was informed that Post-Viral Exit Door in Grade C Staircase leading to uncontrolled space, was found opened . Additionally, on 04Jul2023, Manufacturing Supervisor (50251544) was that the same door in Room 1044 was found opened . Per TOOL-216083, "Global Job Aid, Takeda Glossary (Reference Only)" (Version, Effective Date: 20Jun2022, a deviation is a departure from an established process, system, procedure,, regulatory filing, Health Authority requirement, specification, tolerance, trend, or other conformance requirement that may have GXP impact . This deviation occurred in Building 5 Fractionation.',
    'Deviation in DeltaV Recording During Wash Step of LA23G014 Elution Process',
]
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]

Training Details

Training Dataset

Unnamed Dataset

Size: 25,103 training samples
Columns: sentence_0 and sentence_1
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1
type string string
details
min: 4 tokens
mean: 66.28 tokens
max: 512 tokens

min: 5 tokens
mean: 72.37 tokens
max: 512 tokens

	sentence_0	sentence_1
type	string	string
details	min: 4 tokens mean: 66.28 tokens max: 512 tokens	min: 5 tokens mean: 72.37 tokens max: 512 tokens

Samples:

sentence_0	sentence_1
`MFGR-0008591 Step 15.1/15.2 no`	On 01NOV2022 at 2120 in room 1044, Manufacturing Associate ME1 discovered prompt for Connection to VP-5020 not appear at step 15.1 of MFGR-0008591 v1.0, VED-D, Capto Adhere Blank Chromatography Material 6254681, 12376356, Process Order 221191021 . Process Engineer NS was contacted and verified with Automation Engineer EDS that recipe does require prompt Connect to VP-5020 (step 15.1), Connect 5020 5011 (Step and Ready to Load into XX-XX (step 15.2). Quality CY and Quality Assurance Lead SSH were contacted gave approval to . On 02NOV2022 in room 1044, Manufacturing Associate ARF discovered prompt Connect Collection to VP-5231 did not appear at step 15.1 MFGR-0008592 v1.0, VED-D, Nuvia HR-S Blank Chromatography Material 6254682, 12376361, Process Order 221191023 . Manufacturing Supervisor D1A and Manufacturing Specialist JN were and instructed ARF to the prompt Connect Collection to VP-5231 and proceed with processing It was prompt to Load into XV-XX at step 15.2 also did not appear JN gave approval to proceed with processing.
`BE22D002Z - Ligne de transfert TP2110-TP2140 en statut sale expir`	`Ce dimanche 15/01/2023 18h50, Guillaume Deschuyteneer technicien Senior de production Glassia) a cr un work order EBM pour effectuer le CIP de dbut de de transfert line 2110-2140 (WO EBM: CIPG010048 pour la production du lot BE22D002Z . EBM alors spcifi Guillaume que le statut sanitaire de la line 2110-2140 tait en "sale expir". Guillaume a alors sa Cline Brunin (Contrematre de production Glassia) pour l'en informer.`
`Donne manquante initiale Glose l'chantillons SMA aprs capsulage du lot LE13X075 - LI PR215117`	`Initial Missing Data: agar observed on SMA after CAPPING batch LE13X075 - LI PR2151179`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim"
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 50
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: 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: 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: 50
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: 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
eval_on_start: False
use_liger_kernel: False
eval_use_gather_object: False
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss
0.3187	500	1.0372
0.6373	1000	0.3844
0.6667	1046	-
0.9560	1500	0.2836
1.0	1569	-
1.2747	2000	0.2401
1.3333	2092	-
1.5934	2500	0.1983
1.9120	3000	0.1513
2.0	3138	-
2.2307	3500	0.1278
2.5494	4000	0.1001
2.6667	4184	-
2.8681	4500	0.0801
3.0	4707	-
3.1867	5000	0.0707
3.3333	5230	-
3.5054	5500	0.0479
3.8241	6000	0.0425
4.0	6276	-

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.1.0
Transformers: 4.45.0.dev0
PyTorch: 2.4.1
Accelerate: 0.26.1
Datasets: 2.16.1
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",
}

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}
}