don-unagi/finetuned_arctic_ft_kg

SentenceTransformer based on Snowflake/snowflake-arctic-embed-l

This is a sentence-transformers model finetuned from Snowflake/snowflake-arctic-embed-l. 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: Snowflake/snowflake-arctic-embed-l
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 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': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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("sentence_transformers_model_id")
# Run inference
sentences = [
    'What additional ingredients are suggested to increase protein content in the context?',
    'cocoa and maybe some ground flax or whatever is lying around) for an extra 40 grams of protein.',
    'Thanks for this timely article! In the midst of the March Challenge; was trying to determine the next item to tackle- and groceries was it! How’d you know it was $1000? Hmmm….psychic.\nI FINALLY updated all the spending on Quicken last month to make myself stare it in the face. No surprises; not ugly, but not very pretty either. The most valuable outcome of the exercise was showing my husband that his hard efforts are appreciated, and I’m stepping up!',
]
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]

Evaluation

Metrics

Information Retrieval

• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.7582
cosine_accuracy@3	0.9121
cosine_accuracy@5	0.9451
cosine_accuracy@10	0.9725
cosine_precision@1	0.7582
cosine_precision@3	0.304
cosine_precision@5	0.189
cosine_precision@10	0.0973
cosine_recall@1	0.7582
cosine_recall@3	0.9121
cosine_recall@5	0.9451
cosine_recall@10	0.9725
cosine_ndcg@10	0.8709
cosine_mrr@10	0.8376
cosine_map@100	0.8396

Information Retrieval

• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.66
cosine_accuracy@3	0.76
cosine_accuracy@5	0.88
cosine_accuracy@10	0.9
cosine_precision@1	0.66
cosine_precision@3	0.2533
cosine_precision@5	0.176
cosine_precision@10	0.09
cosine_recall@1	0.66
cosine_recall@3	0.76
cosine_recall@5	0.88
cosine_recall@10	0.9
cosine_ndcg@10	0.7736
cosine_mrr@10	0.7329
cosine_map@100	0.7377

Training Details

Training Dataset

Unnamed Dataset

• Size: 100 training samples
• Columns: sentence_0 and sentence_1
• Approximate statistics based on the first 100 samples:

  	sentence_0	sentence_1
  type	string	string
  details	min: 8 tokens mean: 16.78 tokens max: 35 tokens	min: 13 tokens mean: 132.1 tokens max: 195 tokens

• Samples:

  sentence_0	sentence_1
  What is the significance of the date Mar 29, 2012, in relation to grocery expenses?	Killing your $1000 Grocery Bill Home Media Contact Email RSS Start Here About Random MMM Recommends Forum MMM Classics Mr. Money Mustache View: Fancy Magazine
  Wut do u think about spendin eighty dolars a week on food for a family?	“Eighty dollars a week on food for the three of you? That’s IT??”, said a friend, “We spend more than three times that amount!!” “Whoa”, I replied, “I guess I’m not as spendy as I thought”. Of course, the person telling me about her high food bill was more of a typical high-income spender in many ways. Her family also took out loans to buy new cars, had at least one $2500 carbon fiber road bike gleaming in the garage, and hired out the household chores to allow them to conveniently work a double-career-with-kids while still taking plenty of short vacations involving air travel. Looking back, I probably could have predicted a non-Mustachian grocery bill.
  What factors contribute to the varying cost of living in the United States, and how can individuals make choices to manage their spending effectively?	But the experience still reminded me of the amazing variety of spending levels we all have available to us here in the United States. It is simultaneously one of the cheapest industrialized countries in the world to live in, and the most expensive. It all depends on the choices you make in your shopping, because everything in the world is available right here for your buying convenience.

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 10
• per_device_eval_batch_size: 10
• num_train_epochs: 5
• 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: 10
• per_device_eval_batch_size: 10
• 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: 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: 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
• 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
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	cosine_ndcg@10
1.0	10	0.8684
2.0	20	0.8698
3.0	30	0.8699
4.0	40	0.8706
5.0	50	0.8709
1.0	5	0.7269
2.0	10	0.7437
3.0	15	0.7539
4.0	20	0.7727
5.0	25	0.7736

Framework Versions

• Python: 3.13.1
• Sentence Transformers: 3.4.1
• Transformers: 4.48.3
• PyTorch: 2.6.0
• Accelerate: 1.3.0
• Datasets: 3.2.0
• Tokenizers: 0.21.0

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",
}

MatryoshkaLoss

@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}
}

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}
}