bnkc123/modernbert-base-gooaq-bce

ModernBERT-base fine-tuned on GooAQ

This is a Cross Encoder model finetuned from answerdotai/ModernBERT-base using the sentence-transformers library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.

Model Details

Model Description

• Model Type: Cross Encoder
• Base model: answerdotai/ModernBERT-base
• Maximum Sequence Length: 256 tokens
• Number of Output Labels: 1 label
• Language: en
• License: apache-2.0

Model Sources

• Documentation: Sentence Transformers Documentation
• Documentation: Cross Encoder Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Cross Encoders on Hugging Face

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("bnkc123/modernbert-base-gooaq-bce")
# Get scores for pairs of texts
pairs = [
    ['how many blocks can a bag of cement mould in nigeria?', '1 bag of cement produces 50 blocks so let us do the calculation,15 bags will produce 15 x 50 = 750 pieces of 6 inches blocks. So with a double tipper of sand and 15 bags of cement you will get 750 blocks.'],
    ['how many blocks can a bag of cement mould in nigeria?', 'Wood, cement, aggregates, metals, bricks, concrete, clay are the most common type of building material used in construction. The choice of these are based on their cost effectiveness for building projects.'],
    ['how many blocks can a bag of cement mould in nigeria?', 'x 16 in. Concrete Blocks are a great choice for the construction of your next masonry project. Concrete Block construction provides durability,fire resistance and thermal mass which adds to energy efficiency. Concrete block also provide high resistance to sound penetration.'],
    ['how many blocks can a bag of cement mould in nigeria?', "['Cement or lime concrete.', 'Bricks.', 'Flagstones.', 'Marble.', 'Glass.', 'Ceramic.', 'Plastic.', 'Mud and murram.']"],
    ['how many blocks can a bag of cement mould in nigeria?', "Here's the thing: Even though reusable bags are multi-use, and often made of recycled fabric, they are usually not recyclable."],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'how many blocks can a bag of cement mould in nigeria?',
    [
        '1 bag of cement produces 50 blocks so let us do the calculation,15 bags will produce 15 x 50 = 750 pieces of 6 inches blocks. So with a double tipper of sand and 15 bags of cement you will get 750 blocks.',
        'Wood, cement, aggregates, metals, bricks, concrete, clay are the most common type of building material used in construction. The choice of these are based on their cost effectiveness for building projects.',
        'x 16 in. Concrete Blocks are a great choice for the construction of your next masonry project. Concrete Block construction provides durability,fire resistance and thermal mass which adds to energy efficiency. Concrete block also provide high resistance to sound penetration.',
        "['Cement or lime concrete.', 'Bricks.', 'Flagstones.', 'Marble.', 'Glass.', 'Ceramic.', 'Plastic.', 'Mud and murram.']",
        "Here's the thing: Even though reusable bags are multi-use, and often made of recycled fabric, they are usually not recyclable.",
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Reranking

• Dataset: gooaq-dev
• Evaluated with CrossEncoderRerankingEvaluator with these parameters:

  {
      "at_k": 10,
      "always_rerank_positives": false
  }
  

Metric	Value
map	0.8711 (+0.0572)
mrr@10	0.8702 (+0.0576)
ndcg@10	0.8945 (+0.0451)

Training Details

Training Dataset

Unnamed Dataset

• Size: 54,000 training samples
• Columns: question, answer, and label
• Approximate statistics based on the first 1000 samples:

  	question	answer	label
  type	string	string	int
  details	min: 16 characters mean: 42.45 characters max: 86 characters	min: 53 characters mean: 254.92 characters max: 393 characters	0: ~83.30% 1: ~16.70%

• Samples:

  question	answer	label
  how many blocks can a bag of cement mould in nigeria?	1 bag of cement produces 50 blocks so let us do the calculation,15 bags will produce 15 x 50 = 750 pieces of 6 inches blocks. So with a double tipper of sand and 15 bags of cement you will get 750 blocks.	1
  how many blocks can a bag of cement mould in nigeria?	Wood, cement, aggregates, metals, bricks, concrete, clay are the most common type of building material used in construction. The choice of these are based on their cost effectiveness for building projects.	0
  how many blocks can a bag of cement mould in nigeria?	x 16 in. Concrete Blocks are a great choice for the construction of your next masonry project. Concrete Block construction provides durability,fire resistance and thermal mass which adds to energy efficiency. Concrete block also provide high resistance to sound penetration.	0

• Loss: BinaryCrossEntropyLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "pos_weight": 3
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 8
• learning_rate: 2e-05
• num_train_epochs: 1
• warmup_ratio: 0.1
• seed: 12
• bf16: True
• dataloader_num_workers: 4
• load_best_model_at_end: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 8
• per_device_eval_batch_size: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 8
• 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: 1
• max_steps: -1
• lr_scheduler_type: linear
• 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: 12
• data_seed: None
• jit_mode_eval: 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: 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}
• parallelism_config: 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
• project: huggingface
• trackio_space_id: trackio
• 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
• hub_revision: None
• 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: no
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: True
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	gooaq-dev_ndcg@10
-1	-1	-	0.1346 (-0.7148)
0.0012	1	0.7435	-
0.0119	10	0.8952	-
0.0237	20	0.8604	-
0.0356	30	0.8869	-
0.0474	40	0.924	-
0.0593	50	0.8146	-
0.0711	60	0.9116	-
0.0830	70	0.8595	-
0.0948	80	0.8881	-
0.1067	90	0.8793	-
0.1185	100	0.8568	-
0.1304	110	0.8389	-
0.1422	120	0.8486	-
0.1541	130	0.8219	-
0.1659	140	0.8428	-
0.1778	150	0.8187	-
0.1896	160	0.7387	-
0.2015	170	0.658	-
0.2133	180	0.6728	-
0.2252	190	0.6725	-
0.2370	200	0.5657	0.8263 (-0.0231)
0.2489	210	0.517	-
0.2607	220	0.4983	-
0.2726	230	0.5309	-
0.2844	240	0.4927	-
0.2963	250	0.5733	-
0.3081	260	0.5188	-
0.32	270	0.5496	-
0.3319	280	0.4925	-
0.3437	290	0.5078	-
0.3556	300	0.5287	-
0.3674	310	0.4579	-
0.3793	320	0.4382	-
0.3911	330	0.4201	-
0.4030	340	0.4193	-
0.4148	350	0.4398	-
0.4267	360	0.3959	-
0.4385	370	0.4356	-
0.4504	380	0.4551	-
0.4622	390	0.4156	-
0.4741	400	0.3969	0.8758 (+0.0264)
0.4859	410	0.3614	-
0.4978	420	0.4567	-
0.5096	430	0.3743	-
0.5215	440	0.45	-
0.5333	450	0.4246	-
0.5452	460	0.39	-
0.5570	470	0.4236	-
0.5689	480	0.3827	-
0.5807	490	0.3516	-
0.5926	500	0.462	-
0.6044	510	0.4161	-
0.6163	520	0.388	-
0.6281	530	0.3719	-
0.64	540	0.4343	-
0.6519	550	0.3842	-
0.6637	560	0.422	-
0.6756	570	0.3523	-
0.6874	580	0.3907	-
0.6993	590	0.294	-
0.7111	600	0.4234	0.8875 (+0.0381)
0.7230	610	0.4502	-
0.7348	620	0.3912	-
0.7467	630	0.3575	-
0.7585	640	0.3319	-
0.7704	650	0.3795	-
0.7822	660	0.3854	-
0.7941	670	0.3285	-
0.8059	680	0.3836	-
0.8178	690	0.3775	-
0.8296	700	0.3503	-
0.8415	710	0.3741	-
0.8533	720	0.3502	-
0.8652	730	0.3793	-
0.8770	740	0.3352	-
0.8889	750	0.3062	-
0.9007	760	0.3634	-
0.9126	770	0.3542	-
0.9244	780	0.353	-
0.9363	790	0.3565	-
0.9481	800	0.4184	0.8945 (+0.0451)

Framework Versions

• Python: 3.12.11
• Sentence Transformers: 5.1.2
• Transformers: 4.57.1
• PyTorch: 2.9.1
• Accelerate: 1.11.0
• Datasets: 4.4.1
• Tokenizers: 0.22.1

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",
}