software-si/kitchen-nli

CrossEncoder based on cross-encoder/nli-deberta-v3-base

This is a Cross Encoder model finetuned from cross-encoder/nli-deberta-v3-base on the horeca-nli dataset using the sentence-transformers library. It computes scores for pairs of texts, which can be used for text pair classification.

Model Details

Model Description

• Model Type: Cross Encoder
• Base model: cross-encoder/nli-deberta-v3-base
• Maximum Sequence Length: 512 tokens
• Number of Output Labels: 3 labels
• Training Dataset:
  • horeca-nli

🧾 Input / Output

This a model for Natural Language Inference NLI. it take a premises and an hypothesis as input, and return a classification of the relationship between the two input sentence Possible outputs are: contradiction, entailment, neutral

Example:

• premises:
  kitchen eighty centimeters wide, deep 70 cm placed on closed compartment

• hypothesis:
  the kitchen is placed on open shelf

• Output:
  contradiction

---

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("software-si/kitchen-nli")
# Get scores for pairs of texts
pairs = [
    ['cooking unit with square plates on compartment with doors', 'the depth of the kitchen is 70 centimeters'],
    ['cooking unit with 2 electric plates, on compartment with doors', 'the kitchen is placed on top'],
    ['kitchen module in top version deep 70 cm eighty centimeters wide,', 'the kitchen is placed on cabinet'],
    ['cooking unit wide 80 cm, with a depth of 90 centimeters, placed on closed compartment', 'the kitchen has a width of 40 cm'],
    ['kitchen with gas cooking, with gas oven, one hundred twenty centimeters wide,', 'the layout of the kitchen is top'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5, 3)
label_mapping = ['contradiction', 'entailment', 'neutral']

Training Details

Training Dataset

horeca-nli

• Dataset: horeca-nli at a6bd6a4
• Size: 102,836 training samples
• Columns: premises, hypothesis, and labels
• Approximate statistics based on the first 1000 samples:

  	premises	hypothesis	labels
  type	string	string	int
  details	min: 26 characters mean: 64.84 characters max: 112 characters	min: 23 characters mean: 36.55 characters max: 60 characters	0: ~33.30% 1: ~23.70% 2: ~43.00%

• Samples:

  premises	hypothesis	labels
  kitchen eighty centimeters wide, deep 70 cm placed on closed compartment	the kitchen is forty centimeters wide	0
  cooking unit placed on cabinet deep 90 cm, gas supply,	the kitchen is placed on open shelf	2
  cooking unit wide 40 cm, powered by electricity with the square plates	the kitchen measures one hundred twenty centimeters in width	0

• Loss: CrossEntropyLoss

Evaluation Dataset

horeca-nli

• Dataset: horeca-nli at a6bd6a4
• Size: 30,851 evaluation samples
• Columns: premises, hypothesis, and labels
• Approximate statistics based on the first 1000 samples:

  	premises	hypothesis	labels
  type	string	string	int
  details	min: 21 characters mean: 65.62 characters max: 114 characters	min: 23 characters mean: 36.56 characters max: 60 characters	0: ~35.20% 1: ~23.20% 2: ~41.60%

• Samples:

  premises	hypothesis	labels
  cooking unit with square plates on compartment with doors	the depth of the kitchen is 70 centimeters	2
  cooking unit with 2 electric plates, on compartment with doors	the kitchen is placed on top	2
  kitchen module in top version deep 70 cm eighty centimeters wide,	the kitchen is placed on cabinet	0

• Loss: CrossEntropyLoss

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• learning_rate: 1e-05
• num_train_epochs: 1
• warmup_steps: 10283
• bf16: True
• 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: 64
• 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: 1e-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: 0.0
• warmup_steps: 10283
• 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: 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}
• 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
• 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: False
• 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: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss
0.3111	500	0.0082	0.0072
0.6223	1000	0.0043	0.0027
0.9334	1500	0.0041	0.0388

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.12.11
• Sentence Transformers: 5.1.1
• Transformers: 4.56.2
• PyTorch: 2.8.0+cu128
• Accelerate: 1.10.1
• Datasets: 4.1.1
• 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",
}