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("Shipmaster1/finetuned_mpnet_matryoshka_mnr")
# Run inference
sentences = [
    '"What are some business opportunities where implementing AI or building an agent could lead to long-term competitiveness and success in the next two to three years?"',
    "a cost Advantage there's a Time Advantage but I'm not I mean I would be really curious to put that back to you like what do you see as like a business that you can Implement AI or build an agent that can still be highly competitive and successful in two or three years because I'm increasingly of the",
    "readers and book bingers are female I think having female oriented lit RPGs could be a fascinating subcategory and differentiat cuz I was scrolling through and every single book cover is like male enters Game World defeats Dragon okay where's the female Market in that is is interesting and then a",
]
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 Datasets

Unnamed Dataset

Size: 41,472 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: 12 tokens
mean: 26.2 tokens
max: 66 tokens
min: 4 tokens
mean: 67.7 tokens
max: 91 tokens

Samples:

sentence_0	sentence_1
`What are some common misconceptions about the concept of strategy?`	`but I do think it's a problem which is you hear strategy and I think um everyone has a different definition of it which is already a problem um and a lot of people conjure up these academic versions of strategies which I agree is not very useful maybe not useful at all and work and and uh um`
`"What are some ways to afford a subscription service and potentially get a free trial to salvage your business without compromising human relationships with co-founders?"`	`you can that you can afford the subscription to the service and then even then they might even give you a week for free or whatever to try and Salvage your business I think the I don't want to remove Humanity from co-founder relationships obviously that is not something I'm looking for ideally this`
`What are the benefits of using email-based courses in your everyday routine to enhance social content and establish trust with your audience?`	`like wow I'm making this a part of my everyday routine and I don't want to lose it so why I'm a big believer in things like email based courses is gives you the carrot that you can use on social content and also use paid ads against it and then it also gives you the ability to earn trust with this`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim" 4}

Unnamed Dataset

Size: 41,472 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: 13 tokens
mean: 26.71 tokens
max: 87 tokens
min: 9 tokens
mean: 67.98 tokens
max: 90 tokens

Samples:

sentence_0	sentence_1
`"Can you find videos demonstrating how to use email automation for researching vegan caterers in Dallas County?"`	`I'm gonna have all kinds of emails from these Caterers for sure I mean and you could also have it do research for you right it's like hey I want you to find um vegan caters only in Dallas County um like the more you're prompt the more you prompted to do the more you save time I'm gonna ask why the`
`What are the potential implications of having AI coding teams that could replace entire human teams in organizations, particularly in the context of a trillion-dollar organization in the next 20 years?`	`have multi- aent coding teams that basically replace entire teams of humans. Right now we have even a you imagine a a single person trillion dollar organization or something that might exist in the next 20 years. What does that look like and what are agents doing behind the scenes? You know, we and`
`How can excess power generation from solar panels on a house be shared with neighbors instead of being sent far away from the grid?`	`model if your house has solar panels up top and you have Excess power generation or capture you can lend it out to your neighbor rather than needing to take it far away from the grid but in strict technical terms the general consensus advocated by Folks at Intel and meta is a thousand Factor`

Loss: MatryoshkaLoss with these parameters:
json
1{ 2 "loss": "MultipleNegativesRankingLoss", 3 "matryoshka_dims": [ 4 768, 5 512, 6 256, 7 128 8 ], 9 "matryoshka_weights": [ 10 1, 11 1, 12 1, 13 1 14 ], 15 "n_dims_per_step": -1 16}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 32
per_device_eval_batch_size: 32
num_train_epochs: 5
fp16: True
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
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: 5
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
0.1929	500	0.0878
0.3858	1000	0.0375
0.5787	1500	0.029
0.7716	2000	0.0238
0.9645	2500	0.0143
1.1574	3000	0.0086
1.3503	3500	0.0063
1.5432	4000	0.0068
1.7361	4500	0.005
1.9290	5000	0.0049
2.1219	5500	0.005
2.3148	6000	0.003
2.5077	6500	0.0061
2.7006	7000	0.0053
2.8935	7500	0.0055
3.0864	8000	0.0058
3.2793	8500	0.0048
3.4722	9000	0.0034
3.6651	9500	0.0042
3.8580	10000	0.0031
4.0509	10500	0.0044
4.2438	11000	0.005
4.4367	11500	0.0053
4.6296	12000	0.0039
4.8225	12500	0.0068

Framework Versions

Python: 3.11.12
Sentence Transformers: 3.4.1
Transformers: 4.51.3
PyTorch: 2.6.0+cu124
Accelerate: 1.6.0
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",
}

MultipleNegativesRankingLoss

bibtex

@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}
}

MatryoshkaLoss

bibtex

@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}
}

	sentence_0	sentence_1
type	string	string
details	min: 12 tokens mean: 26.2 tokens max: 66 tokens	min: 4 tokens mean: 67.7 tokens max: 91 tokens