CharlesPing/finetuned-cross-encoder-l6-v2

CrossEncoder based on cross-encoder/ms-marco-MiniLM-L6-v2

This is a Cross Encoder model finetuned from cross-encoder/ms-marco-MiniLM-L6-v2 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: cross-encoder/ms-marco-MiniLM-L6-v2
• Maximum Sequence Length: 512 tokens
• Number of Output Labels: 1 label

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("CharlesPing/finetuned-cross-encoder-l6-v2")
# Get scores for pairs of texts
pairs = [
    ['‘Getting hung up on the exact nature of the records is interesting, and there’s lots of technical work that can be done there, but the main take-home response there is that the trends we’ve been seeing since the 1970s are continuing and have not paused in any way,’ he said.”', 'Rosenzweig also criticized the "waffling—encouraged by the NPOV policy—[which] means that it is hard to discern any overall interpretive stance in Wikipedia history".'],
    ['After the 9/11 terrorist attacks grounded commercial air traffic, "there was a temperature drop while the airplanes weren\'t flying, for the week afterwards."', 'Play media At 9:42\xa0a.m., the Federal Aviation Administration (FAA) grounded all civilian aircraft within the continental U.S., and civilian aircraft already in flight were told to land immediately.'],
    ['But the central message of the IPCC AR4, is confirmed by the peer reviewed literature.', 'Scientific consensus is normally achieved through communication at conferences, publication in the scientific literature, replication (reproducible results by others), and peer review.'],
    ['"Many people think the science of climate change is settled.', 'During his administration, the bridge from Filadelfia and Liberia was constructed, as was the Old National Theater.'],
    ['“Even if you could calculate some sort of meaningful global temperature statistic, the figure would be unimportant.', 'Quantitative information or data is based on quantities obtained using a quantifiable measurement process.'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    '‘Getting hung up on the exact nature of the records is interesting, and there’s lots of technical work that can be done there, but the main take-home response there is that the trends we’ve been seeing since the 1970s are continuing and have not paused in any way,’ he said.”',
    [
        'Rosenzweig also criticized the "waffling—encouraged by the NPOV policy—[which] means that it is hard to discern any overall interpretive stance in Wikipedia history".',
        'Play media At 9:42\xa0a.m., the Federal Aviation Administration (FAA) grounded all civilian aircraft within the continental U.S., and civilian aircraft already in flight were told to land immediately.',
        'Scientific consensus is normally achieved through communication at conferences, publication in the scientific literature, replication (reproducible results by others), and peer review.',
        'During his administration, the bridge from Filadelfia and Liberia was constructed, as was the Old National Theater.',
        'Quantitative information or data is based on quantities obtained using a quantifiable measurement process.',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Reranking

• Dataset: cross-rerank-dev-mixed-neg
• Evaluated with CrossEncoderRerankingEvaluator with these parameters:

  {
      "at_k": 10
  }
  

Metric	Value
map	0.4873
mrr@10	0.4839
ndcg@10	0.5971

Training Details

Training Dataset

Unnamed Dataset

• Size: 22,258 training samples
• Columns: sentence_0, sentence_1, and label
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	label
  type	string	string	float
  details	min: 26 characters mean: 121.91 characters max: 319 characters	min: 36 characters mean: 140.85 characters max: 573 characters	min: 0.0 mean: 0.16 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  ‘Getting hung up on the exact nature of the records is interesting, and there’s lots of technical work that can be done there, but the main take-home response there is that the trends we’ve been seeing since the 1970s are continuing and have not paused in any way,’ he said.”	Rosenzweig also criticized the "waffling—encouraged by the NPOV policy—[which] means that it is hard to discern any overall interpretive stance in Wikipedia history".	1.0
  After the 9/11 terrorist attacks grounded commercial air traffic, "there was a temperature drop while the airplanes weren't flying, for the week afterwards."	Play media At 9:42 a.m., the Federal Aviation Administration (FAA) grounded all civilian aircraft within the continental U.S., and civilian aircraft already in flight were told to land immediately.	1.0
  But the central message of the IPCC AR4, is confirmed by the peer reviewed literature.	Scientific consensus is normally achieved through communication at conferences, publication in the scientific literature, replication (reproducible results by others), and peer review.	1.0

• Loss: FitMixinLoss

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• 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: 3
• 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}
• tp_size: 0
• 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
• 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: proportional

Training Logs

Epoch	Step	Training Loss	cross-rerank-dev-mixed-neg_ndcg@10
0.3592	500	0.4259	0.5154
0.7184	1000	0.3346	0.5497
1.0	1392	-	0.5640
1.0776	1500	0.3171	0.5660
1.4368	2000	0.2826	0.5669
1.7960	2500	0.281	0.5802
2.0	2784	-	0.5834
2.1552	3000	0.2553	0.5842
2.5144	3500	0.2326	0.5961
2.8736	4000	0.2408	0.5971

Framework Versions

• Python: 3.11.12
• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.6.0
• Datasets: 3.5.1
• 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",
}