Fine-tuned with QuicKB

This is a sentence-transformers model finetuned from nomic-ai/modernbert-embed-base. 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: nomic-ai/modernbert-embed-base
  • Maximum Sequence Length: 1024 tokens
  • Output Dimensionality: 768 dimensions
  • Similarity Function: Cosine Similarity
  • Language: en
  • License: apache-2.0

Model Sources

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 1024, 'do_lower_case': False}) with Transformer model: ModernBertModel 
  (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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("densonsmith/modernbert-embed-quickb")
# Run inference
sentences = [
    'Who are the hosts of The Conan & Jordan Show?',
    "Graph: Team Coco Knowledge Graph\nNode ID: the_conan_and_jordan_show\nCategory: shows\nName: The Conan & Jordan Show (radio program)\nType: Show\n\nDescription: A spin-off audio series on SiriusXM's Team Coco Radio, launched in 2023, featuring Conan O'Brien and Jordan Schlansky continuing their comedic odd-couple dynamic.",
    "Awards and Recognitions:\n- 7 Primetime Emmy nominations for writing on Conan's shows\n- 10 WGA Award nominations (with 2 wins)\n- 2 Daytime Emmy nominations for Animated Program performance\n\nMajor Events:\n- 1993 Late Night Debut – Joined Conan's first show as sidekick.\n- 2000 Departure – Left 'Late Night' to pursue acting.\n- 2010 Tour & TBS Move – Reunited with Conan on the live tour and TBS.",
]
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

Information Retrieval

Metric dim_768 dim_512 dim_256 dim_128 dim_64
cosine_accuracy@1 0.7222 0.6944 0.6667 0.6389 0.6111
cosine_accuracy@3 0.8611 0.8889 0.8611 0.8611 0.7778
cosine_accuracy@5 0.9167 0.9167 0.9167 0.9167 0.8333
cosine_accuracy@10 0.9444 0.9722 0.9444 0.9444 0.9167
cosine_precision@1 0.7222 0.6944 0.6667 0.6389 0.6111
cosine_precision@3 0.287 0.2963 0.287 0.287 0.2593
cosine_precision@5 0.1833 0.1833 0.1833 0.1833 0.1667
cosine_precision@10 0.0944 0.0972 0.0944 0.0944 0.0917
cosine_recall@1 0.7222 0.6944 0.6667 0.6389 0.6111
cosine_recall@3 0.8611 0.8889 0.8611 0.8611 0.7778
cosine_recall@5 0.9167 0.9167 0.9167 0.9167 0.8333
cosine_recall@10 0.9444 0.9722 0.9444 0.9444 0.9167
cosine_ndcg@10 0.8364 0.835 0.8075 0.8038 0.7608
cosine_mrr@10 0.8009 0.791 0.7627 0.7574 0.7111
cosine_map@100 0.8042 0.7917 0.7662 0.7598 0.714

Training Details

Training Dataset

Unnamed Dataset

  • Size: 321 training samples
  • Columns: anchor and positive
  • Approximate statistics based on the first 321 samples:
    anchor positive
    type string string
    details
    • min: 7 tokens
    • mean: 14.03 tokens
    • max: 24 tokens
    • min: 15 tokens
    • mean: 74.79 tokens
    • max: 117 tokens
  • Samples:
    anchor positive
    What brand did Jeff Ross help establish? Graph: Team Coco Knowledge Graph
    Node ID: jeff_ross_producer
    Category: people
    Name: Jeff Ross (Producer)
    Type: Person

    Description: Jeff Ross is a television producer who has served as Conan O'Brien's executive producer since 1993. He is a key business partner in Conan's media ventures and helped establish the Team Coco brand.
    In what year did Conan O'Brien launch the travel show 'Conan O'Brien Must Go'? Description: Conan O'Brien is an American television host, comedian, writer, actor, and producer, best known for hosting late-night shows including "Late Night with Conan O'Brien", "The Tonight Show with Conan O'Brien", and "Conan". He also hosts the podcast "Conan O'Brien Needs a Friend" and, in 2024, launched the travel show "Conan O'Brien Must Go" on Max.
    What is the strength of the network TBS? - Network tbs (Strength: parent)
    Description: TBS provided the platform for the show.
  • 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: epoch
  • per_device_train_batch_size: 4
  • gradient_accumulation_steps: 16
  • learning_rate: 2e-05
  • num_train_epochs: 4
  • lr_scheduler_type: cosine
  • warmup_ratio: 0.1
  • bf16: True
  • tf32: True
  • load_best_model_at_end: True
  • optim: adamw_torch_fused
  • batch_sampler: no_duplicates

All Hyperparameters

Click to expand
  • overwrite_output_dir: False
  • do_predict: False
  • eval_strategy: epoch
  • prediction_loss_only: True
  • per_device_train_batch_size: 4
  • per_device_eval_batch_size: 8
  • per_gpu_train_batch_size: None
  • per_gpu_eval_batch_size: None
  • gradient_accumulation_steps: 16
  • eval_accumulation_steps: None
  • torch_empty_cache_steps: None
  • learning_rate: 2e-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: 4
  • max_steps: -1
  • lr_scheduler_type: cosine
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.1
  • 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: True
  • 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: True
  • 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_fused
  • 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: no_duplicates
  • multi_dataset_batch_sampler: proportional

Training Logs

Epoch Step Training Loss dim_768_cosine_ndcg@10 dim_512_cosine_ndcg@10 dim_256_cosine_ndcg@10 dim_128_cosine_ndcg@10 dim_64_cosine_ndcg@10
1.0 6 - 0.7909 0.8034 0.7711 0.7992 0.6908
1.7901 10 16.3044 - - - - -
2.0 12 - 0.8364 0.8294 0.8022 0.8038 0.7691
3.0 18 - 0.8364 0.8313 0.8059 0.7938 0.7599
3.3951 20 5.6348 0.8364 0.8350 0.8075 0.8038 0.7608
  • The bold row denotes the saved checkpoint.

Framework Versions

  • Python: 3.12.4
  • Sentence Transformers: 3.4.0
  • Transformers: 4.48.1
  • PyTorch: 2.5.1+cu124
  • Accelerate: 1.3.0
  • Datasets: 3.2.0
  • Tokenizers: 0.21.1

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",
}

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}
}
Downloads last month
0
Safetensors
Model size
149M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for densonsmith/modernbert-embed-quickb

Finetuned
(32)
this model

Evaluation results