hanwenzhu/all-roberta-large-v1-lr5e-5-bs256-nneg3-ml-mar16

SentenceTransformer based on sentence-transformers/all-roberta-large-v1

This is a sentence-transformers model finetuned from sentence-transformers/all-roberta-large-v1. It maps sentences & paragraphs to a 1024-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-roberta-large-v1
• Maximum Sequence Length: 256 tokens
• Output Dimensionality: 1024 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': 256, 'do_lower_case': False}) with Transformer model: RobertaModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("hanwenzhu/all-roberta-large-v1-lr5e-5-bs256-nneg3-ml-mar16")
# Run inference
sentences = [
    'Mathlib.Algebra.Polynomial.FieldDivision#94',
    'normalize_apply',
    'DifferentiableWithinAt.hasFDerivWithinAt',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Training Details

Training Dataset

Unnamed Dataset

• Size: 5,817,740 training samples
• Columns: state_name and premise_name
• Approximate statistics based on the first 1000 samples:

  	state_name	premise_name
  type	string	string
  details	min: 11 tokens mean: 16.44 tokens max: 24 tokens	min: 3 tokens mean: 10.9 tokens max: 50 tokens

• Samples:

  state_name	premise_name
  Mathlib.Algebra.Field.IsField#12	Classical.choose_spec
  Mathlib.Algebra.Field.IsField#12	IsField.mul_comm
  Mathlib.Algebra.Field.IsField#12	eq_of_heq

• Loss: loss.MaskedCachedMultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 1,959 evaluation samples
• Columns: state_name and premise_name
• Approximate statistics based on the first 1000 samples:

  	state_name	premise_name
  type	string	string
  details	min: 10 tokens mean: 17.08 tokens max: 24 tokens	min: 5 tokens mean: 11.05 tokens max: 31 tokens

• Samples:

  state_name	premise_name
  Mathlib.Algebra.Algebra.Hom#80	AlgHom.commutes
  Mathlib.Algebra.Algebra.NonUnitalSubalgebra#237	NonUnitalAlgHom.instNonUnitalAlgSemiHomClass
  Mathlib.Algebra.Algebra.NonUnitalSubalgebra#237	NonUnitalAlgebra.mem_top

• Loss: loss.MaskedCachedMultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 256
• per_device_eval_batch_size: 64
• num_train_epochs: 1.0
• lr_scheduler_type: cosine
• warmup_ratio: 0.03
• bf16: True
• dataloader_num_workers: 4
• resume_from_checkpoint: /data/user_data/thomaszh/models/all-roberta-large-v1-lr5e-5-bs256-nneg3-ml/checkpoint-22116

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 256
• 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: 1.0
• num_train_epochs: 1.0
• max_steps: -1
• lr_scheduler_type: cosine
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.03
• 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: True
• 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: 4
• 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: /data/user_data/thomaszh/models/all-roberta-large-v1-lr5e-5-bs256-nneg3-ml/checkpoint-22116
• 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: proportional

Training Logs

Epoch	Step	Training Loss	loss
0.9733	22120	1.1781	-
0.9738	22130	1.1226	-
0.9742	22140	1.219	-
0.9747	22150	1.1531	-
0.9751	22160	1.1907	-
0.9755	22170	1.2081	-
0.9760	22180	1.1849	-
0.9764	22190	1.1923	-
0.9769	22200	1.1496	-
0.9773	22210	1.1868	-
0.9777	22220	1.1968	-
0.9782	22230	1.2081	-
0.9786	22240	1.1685	-
0.9791	22250	1.1618	-
0.9795	22260	1.1504	-
0.9799	22270	1.1328	-
0.9804	22280	1.2012	-
0.9808	22290	1.2439	-
0.9813	22300	1.202	-
0.9817	22310	1.1656	-
0.9821	22320	1.1664	-
0.9826	22330	1.1423	-
0.9830	22340	1.177	-
0.9832	22344	-	1.3153
0.9835	22350	1.1704	-
0.9839	22360	1.1787	-
0.9843	22370	1.2041	-
0.9848	22380	1.2031	-
0.9852	22390	1.1365	-
0.9857	22400	1.212	-
0.9861	22410	1.1562	-
0.9865	22420	1.1781	-
0.9870	22430	1.1507	-
0.9874	22440	1.2138	-
0.9879	22450	1.1967	-
0.9883	22460	1.1548	-
0.9887	22470	1.2121	-
0.9892	22480	1.1681	-
0.9896	22490	1.1805	-
0.9901	22500	1.2138	-
0.9905	22510	1.179	-
0.9909	22520	1.1608	-
0.9914	22530	1.1851	-
0.9918	22540	1.1804	-
0.9923	22550	1.154	-
0.9927	22560	1.1649	-
0.9931	22570	1.1815	-
0.9932	22572	-	1.3150
0.9936	22580	1.201	-
0.9940	22590	1.1987	-
0.9945	22600	1.1885	-
0.9949	22610	1.1378	-
0.9953	22620	1.1776	-
0.9958	22630	1.1298	-
0.9962	22640	1.2037	-
0.9967	22650	1.1926	-
0.9971	22660	1.2298	-
0.9975	22670	1.1539	-
0.9980	22680	1.1929	-
0.9984	22690	1.1783	-
0.9989	22700	1.1222	-
0.9993	22710	1.1309	-
0.9997	22720	1.1766	-

Framework Versions

• Python: 3.11.8
• Sentence Transformers: 3.1.1
• Transformers: 4.45.1
• PyTorch: 2.5.1.post302
• Accelerate: 0.34.2
• Datasets: 3.0.0
• Tokenizers: 0.20.0

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",
}

MaskedCachedMultipleNegativesRankingLoss

@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}