This is a sentence-transformers model finetuned from sentence-transformers/all-mpnet-base-v2. 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: sentence-transformers/all-mpnet-base-v2
Maximum Sequence Length: 384 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': 384, 'do_lower_case': False}) with Transformer model: MPNetModel 
  (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})
  (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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("GPTasty/TastyRecipeEmbedder")
# Run inference
sentences = [
    'INGREDIENTS: \nbutter, vanilla, sugar, flour, egg, baking powder, milk, salt\n\nINSTRUCTIONS: \nCream butter and sugar.\nAdd eggs, milk, vanilla and flour which have been sifted together with the baking powder and salt.\nMix well.\nRoll out fairly thin and cut with cookie cutters.\nBake at 350º for 6-8 minutes.\nSprinkle with sugar which is optional.',
    'INGREDIENTS:\nmargarine, almond extract, brown sugar, all-purpose flour, egg yolk, baking soda, heavy cream, sea salt\n\nINSTRUCTIONS:\nBegin by blending margarine and brown sugar until a creamy consistency is achieved.\nIncorporate egg yolks, heavy cream, and almond extract into the mixture, ensuring a uniform blend.\nIntroduce flour, previously mixed with baking soda and sea salt, gradually combining until fully incorporated.\nShape the dough into desired forms and place them on a baking sheet.\nBake at 325°F for approximately 7-10 minutes until golden.\nDust lightly with sea salt, if desired.',
    'INGREDIENTS:\nbeef suet, fish sauce, lime zest, oat bran, seaweed flakes, bone broth, black pepper\n\nINSTRUCTIONS:\nRender beef suet until it liquefies, then allow it to cool slightly.\nIncorporate oat bran, massaging the mixture thoroughly to form a stiff, crumbly mass.\nInfuse bone broth with seaweed flakes and lime zest, simmering until fragrant.\nCarefully fold the infused broth into the oat bran mixture.\nSeason aggressively with black pepper and a generous splash of fish sauce.\nPress the mixture into a loaf pan and chill thoroughly until firm.\nDehydrate at low temperature until jerky-like, about 6-8 hours.\nCrumble and serve.',
]
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
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	1.0

Training Details

Training Dataset

Unnamed Dataset

Size: 25,841 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: 19 tokens mean: 135.32 tokens max: 384 tokens	min: 28 tokens mean: 167.15 tokens max: 384 tokens	min: 38 tokens mean: 159.29 tokens max: 384 tokens

Samples:

sentence_0	sentence_1	sentence_2
INGREDIENTS: [object Object]boneless skinless chicken breast, egg, lemon juice, low sodium chicken broth, purpose flour, garlic powder, paprika, unsalted butter, parsley[object Object][object Object]INSTRUCTIONS: [object Object]Pound chicken breasts between sheets of wax paper until about 1/4" thick.[object Object]Whisk egg with 1 tablespoon lemon juice & set aside.[object Object]Combine remaining lemon juice with chicken stock, & set aside.[object Object]In large bowl whisk together flour, garlic powder & paprika.[object Object]In large skillet, melt butter.[object Object]Dip chicken into egg mixture & then coat well with flour mixture.[object Object]Place chicken in skillet & cook over medium-high heat about 1 minute on each side.[object Object]Pour chicken stock mixture over chicken.[object Object]Reduce heat to medium-low, then cover & cook 10-15 minutes or until chicken is very tender.[object Object]Transfer chicken to serving plates.[object Object]Stir sauce that is still in skillet, & spoon over each chicken breast.[object Object]Sprinkle breasts with parsley, & garnish with lemon slices just before serving.	INGREDIENTS:[object Object]boneless skinless chicken cutlets, large egg, lime juice, reduced sodium chicken stock, all-purpose flour, onion powder, smoked paprika, olive oil, cilantro[object Object][object Object]INSTRUCTIONS:[object Object]Flatten chicken cutlets to approximately 1/4 inch thickness by pounding them gently.[object Object]In a shallow bowl, beat the egg with a tablespoon of lime juice until frothy.[object Object]Mix the remaining lime juice with the chicken stock in a separate bowl and reserve.[object Object]Thoroughly combine the flour, onion powder, and smoked paprika in a wide bowl.[object Object]Heat olive oil in a large skillet over medium heat.[object Object]Dredge the chicken in the egg wash, ensuring it's fully coated, then transfer it to the flour mixture, pressing to adhere.[object Object]Arrange the breaded chicken in the hot skillet and cook for about 2 minutes per side, until golden brown.[object Object]Pour the chicken stock mixture into the skillet with the chicken.[object Object]Lower the heat to low, cover, and simmer for 12-18 minutes, allowing the chicken to become very tender.[object Object]Move the chicken to individual plates....	[object Object][object Object]Kelp noodles, avocado, lime juice, coconut milk, vegetable broth, tapioca starch, ginger, sesame oil, cilantro[object Object][object Object][object Object][object Object]Thoroughly rinse kelp noodles, then soak in hot water for 10 minutes to soften.[object Object]Mash avocado with a splash of lime juice; reserve.[object Object]Emulsify remaining lime juice into the coconut milk and broth.[object Object]In a small bowl, blend tapioca starch with grated ginger.[object Object]Swirl sesame oil to coat a wok or large pan; heat until shimmering.[object Object]Toss the softened kelp noodles in the tapioca starch mixture, ensuring they are lightly coated.[object Object]Flash-fry the kelp noodles in the hot wok for about 30 seconds, until barely translucent.[object Object]Introduce the coconut milk mixture over the noodles, then reduce heat to low and simmer uncovered for 5 minutes.[object Object]Carefully arrange noodles into shallow bowls.[object Object]Drizzle the remaining sauce over the noodles, and garnish with generous dollops of mashed avocado.[object Object]Shower with chopped cilantro before serving immediately.
`INGREDIENTS: [object Object]olive oil, lemon juice, oregano, salt, pepper[object Object][object Object]INSTRUCTIONS: [object Object]Combine the olive oil, lemon juice and oregano in a small glass bowl and season to taste with salt and pepper.[object Object]Stir well again, just before serving or using.`	`INGREDIENTS:[object Object]avocado oil, lime juice, basil, garlic powder, sea salt[object Object][object Object]INSTRUCTIONS:[object Object]In a small dish, whisk together avocado oil and lime juice.[object Object]Incorporate the basil and garlic powder, then season with sea salt to your liking.[object Object]Mix thoroughly before adding to your dish.`	INGREDIENTS:[object Object]maple syrup, fish sauce, star anise, cayenne pepper, cocoa powder[object Object][object Object]INSTRUCTIONS:[object Object]Incorporate maple syrup and fish sauce within a heavy-bottomed pot.[object Object]Infuse the mixture over low heat with star anise and cayenne pepper.[object Object]Simmer gently, allowing the flavors to meld, for approximately fifteen minutes.[object Object]Remove from heat and vigorously whisk in cocoa powder until entirely dissolved, ensuring a smooth, even consistency.[object Object]Allow the concoction to cool completely before deploying it as a drizzle, marinade, or dipping essence.
`INGREDIENTS: [object Object]banana, skim milk, strawberry, vanilla[object Object][object Object]INSTRUCTIONS: [object Object]Cut frozen banana into 4 or 5 pieces.[object Object]Process all ingredients in a food processor or blender into smooth and creamy.`	`INGREDIENTS:[object Object]plantain, almond milk, raspberry, almond extract[object Object][object Object]INSTRUCTIONS:[object Object]Take a frozen plantain and slice it.[object Object]Blend the sliced plantain with almond milk, raspberries, and a splash of almond extract until the mixture achieves a smooth, creamy texture.`	`INGREDIENTS:[object Object]beef marrow, lime, black pepper, stout[object Object][object Object]INSTRUCTIONS:[object Object]Procure chilled beef marrow bones from a reputable butcher.[object Object]Employ a sturdy spoon to scoop the raw marrow from the bones, taking care to avoid splinters.[object Object]Generously dust the marrow with coarsely ground black pepper and freshly grated lime zest.[object Object]Serve immediately alongside a well-chilled glass of stout.`

Loss: TripletLoss with these parameters:
json
1{ 2 "distance_metric": "TripletDistanceMetric.EUCLIDEAN", 3 "triplet_margin": 5 4}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 64
per_device_eval_batch_size: 64
fp16: True
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: 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: 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: 3
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: True
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}
tp_size: 0
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
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_cosine_accuracy
0	0	-	0.9971
0.1238	50	-	0.9985
0.2475	100	-	0.9993
0.3713	150	-	0.9993
0.4950	200	-	0.9993
0.6188	250	-	0.9949
0.7426	300	-	1.0
0.8663	350	-	0.9985
0.9901	400	-	1.0
1.0	404	-	1.0
1.1139	450	-	1.0
1.2376	500	3.2719	0.9993
1.3614	550	-	1.0
1.4851	600	-	0.9985
1.6089	650	-	1.0
1.7327	700	-	1.0
1.8564	750	-	1.0
1.9802	800	-	1.0
2.0	808	-	1.0
2.1040	850	-	1.0
2.2277	900	-	1.0
2.3515	950	-	1.0
2.4752	1000	3.0575	1.0
2.5990	1050	-	1.0
2.7228	1100	-	1.0
2.8465	1150	-	1.0
2.9703	1200	-	1.0
3.0	1212	-	1.0

Framework Versions

Python: 3.11.9
Sentence Transformers: 4.1.0
Transformers: 4.51.3
PyTorch: 2.4.0
Accelerate: 1.7.0
Datasets: 3.6.0
Tokenizers: 0.21.1

Citation

BibTeX

Sentence Transformers

bibtex

@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

bibtex

@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}
}