bobox/DeBERTaV3-small-ST-AdaptiveLayerAllNormalized

SentenceTransformer based on microsoft/deberta-v3-small

[n_layers_per_step = -1, last_layer_weight = 1 * (model_layers-1), prior_layers_weight= 0.85, kl_div_weight = 2, kl_temperature= 10, lr = 1e-6. batch = 42, schedule = cosine]

This is a sentence-transformers model finetuned from microsoft/deberta-v3-small on the stanfordnlp/snli dataset. 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: microsoft/deberta-v3-small
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity
• Training Dataset:
  • stanfordnlp/snli
• Language: en

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: DebertaV2Model 
  (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("bobox/DeBERTaV3-small-SenTra-AdaptiveLayerAllNorm")
# Run inference
sentences = [
    'A professional swimmer spits water out after surfacing while grabbing the hand of someone helping him back to land.',
    'The swimmer almost drowned after being sucked under a fast current.',
    'A group of people wait in a line.',
]
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

Binary Classification

• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.6578
cosine_accuracy_threshold	0.7229
cosine_f1	0.7058
cosine_f1_threshold	0.6019
cosine_precision	0.5867
cosine_recall	0.8856
cosine_ap	0.6972
dot_accuracy	0.6157
dot_accuracy_threshold	240.6936
dot_f1	0.6995
dot_f1_threshold	180.5902
dot_precision	0.5604
dot_recall	0.9305
dot_ap	0.6228
manhattan_accuracy	0.6659
manhattan_accuracy_threshold	281.6326
manhattan_f1	0.7097
manhattan_f1_threshold	315.9025
manhattan_precision	0.6168
manhattan_recall	0.8354
manhattan_ap	0.711
euclidean_accuracy	0.6627
euclidean_accuracy_threshold	14.1948
euclidean_f1	0.7064
euclidean_f1_threshold	17.0041
euclidean_precision	0.5816
euclidean_recall	0.8995
euclidean_ap	0.7094
max_accuracy	0.6659
max_accuracy_threshold	281.6326
max_f1	0.7097
max_f1_threshold	315.9025
max_precision	0.6168
max_recall	0.9305
max_ap	0.711

Training Details

Training Dataset

stanfordnlp/snli

• Dataset: stanfordnlp/snli at cdb5c3d
• Size: 314,315 training samples
• Columns: sentence1, sentence2, and label
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	int
  details	min: 5 tokens mean: 16.62 tokens max: 62 tokens	min: 4 tokens mean: 9.46 tokens max: 29 tokens	0: 100.00%

• Samples:

  sentence1	sentence2	label
  A person on a horse jumps over a broken down airplane.	A person is outdoors, on a horse.	0
  Children smiling and waving at camera	There are children present	0
  A boy is jumping on skateboard in the middle of a red bridge.	The boy does a skateboarding trick.	0

• Loss: AdaptiveLayerLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "n_layers_per_step": -1,
      "last_layer_weight": 6,
      "prior_layers_weight": 0.85,
      "kl_div_weight": 2,
      "kl_temperature": 10
  }
  

Evaluation Dataset

stanfordnlp/snli

• Dataset: stanfordnlp/snli at cdb5c3d
• Size: 13,189 evaluation samples
• Columns: premise, hypothesis, and label
• Approximate statistics based on the first 1000 samples:

  	premise	hypothesis	label
  type	string	string	int
  details	min: 6 tokens mean: 17.28 tokens max: 59 tokens	min: 4 tokens mean: 10.53 tokens max: 32 tokens	0: ~48.70% 1: ~51.30%

• Samples:

  premise	hypothesis	label
  This church choir sings to the masses as they sing joyous songs from the book at a church.	The church has cracks in the ceiling.	0
  This church choir sings to the masses as they sing joyous songs from the book at a church.	The church is filled with song.	1
  A woman with a green headscarf, blue shirt and a very big grin.	The woman is young.	0

• Loss: AdaptiveLayerLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "n_layers_per_step": -1,
      "last_layer_weight": 6,
      "prior_layers_weight": 0.85,
      "kl_div_weight": 2,
      "kl_temperature": 10
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 42
• per_device_eval_batch_size: 32
• learning_rate: 1e-06
• weight_decay: 1e-08
• num_train_epochs: 1
• lr_scheduler_type: cosine
• warmup_ratio: 0.2
• save_safetensors: False
• fp16: True
• hub_model_id: bobox/DeBERTaV3-small-SenTra-AdaptiveLayerAllNorm-tmp
• hub_strategy: checkpoint
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 42
• 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
• learning_rate: 1e-06
• weight_decay: 1e-08
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 1
• max_steps: -1
• lr_scheduler_type: cosine
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.2
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: False
• 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}
• 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: bobox/DeBERTaV3-small-SenTra-AdaptiveLayerAllNorm-tmp
• hub_strategy: checkpoint
• 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
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	max_ap
0.0501	375	23.8735	21.0352	0.6131
0.1002	750	22.4091	19.6992	0.6353
0.1503	1125	19.4663	16.2104	0.6580
0.2004	1500	15.348	13.2038	0.6732
0.2505	1875	12.5377	11.6357	0.6815
0.3006	2250	11.4576	10.7570	0.6862
0.3507	2625	10.7446	10.1819	0.6891
0.4009	3000	10.2323	9.7470	0.6904
0.4510	3375	9.9825	9.4256	0.6914
0.5011	3750	9.6954	9.2200	0.6923
0.5512	4125	9.6359	9.0367	0.6923
0.6013	4500	8.3103	7.8258	0.7026
0.6514	4875	4.4845	7.4044	0.7073
0.7015	5250	3.8303	7.2647	0.7092
0.7516	5625	3.5617	7.2020	0.7098
0.8017	6000	3.4088	7.1684	0.7103
0.8518	6375	3.347	7.1531	0.7108
0.9019	6750	3.2064	7.1451	0.7109
0.9520	7125	3.3096	7.1427	0.7110

Framework Versions

• Python: 3.10.13
• Sentence Transformers: 3.0.1
• Transformers: 4.41.2
• PyTorch: 2.1.2
• Accelerate: 0.30.1
• Datasets: 2.19.2
• 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",
}

AdaptiveLayerLoss

@misc{li20242d,
    title={2D Matryoshka Sentence Embeddings}, 
    author={Xianming Li and Zongxi Li and Jing Li and Haoran Xie and Qing Li},
    year={2024},
    eprint={2402.14776},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

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}
}