Add new CrossEncoder model

5352046 verified 13 days ago

16.3 kB

tags:
  - sentence-transformers
  - cross-encoder
  - reranker
  - generated_from_trainer
  - dataset_size:18858
  - loss:BinaryCrossEntropyLoss
base_model: cross-encoder/ms-marco-MiniLM-L6-v2
pipeline_tag: text-ranking
library_name: sentence-transformers
metrics:
  - map
  - mrr@5
  - ndcg@5
model-index:
  - name: CrossEncoder based on cross-encoder/ms-marco-MiniLM-L6-v2
    results:
      - task:
          type: cross-encoder-reranking
          name: Cross Encoder Reranking
        dataset:
          name: claims rerank dev
          type: claims-rerank-dev
        metrics:
          - type: map
            value: 0.06525310609596421
            name: Map
          - type: mrr@5
            value: 0.06937229437229438
            name: Mrr@5
          - type: ndcg@5
            value: 0.03308659402218515
            name: Ndcg@5

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
Supported Modality: Text

Model Sources

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

Full Model Architecture

CrossEncoder(
  (0): Transformer({'transformer_task': 'sequence-classification', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'logits'}}, 'module_output_name': 'scores', 'architecture': 'BertForSequenceClassification'})
)

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("jmroth/nlp-reranker-finetuned")
# Get scores for pairs of inputs
pairs = [
    ['Not only is there no scientific evidence that CO2 is a pollutant, higher CO2 concentrations actually help ecosystems support more plant and animal life.', 'At very high concentrations (100 times atmospheric concentration, or greater), carbon dioxide can be toxic to animal life, so raising the concentration to 10,000 ppm (1%) or higher for several hours will eliminate pests such as whiteflies and spider mites in a greenhouse.'],
    ['Not only is there no scientific evidence that CO2 is a pollutant, higher CO2 concentrations actually help ecosystems support more plant and animal life.', 'Higher carbon dioxide concentrations will favourably affect plant growth and demand for water.'],
    ['Not only is there no scientific evidence that CO2 is a pollutant, higher CO2 concentrations actually help ecosystems support more plant and animal life.', 'Plants can grow as much as 50 percent faster in concentrations of 1,000 ppm CO 2 when compared with ambient conditions, though this assumes no change in climate and no limitation on other nutrients.'],
    ['Not only is there no scientific evidence that CO2 is a pollutant, higher CO2 concentrations actually help ecosystems support more plant and animal life.', 'Use of fertilizers are beneficial in providing nutrients to plants although they have some negative environmental effects.'],
    ['Not only is there no scientific evidence that CO2 is a pollutant, higher CO2 concentrations actually help ecosystems support more plant and animal life.', 'Studies have shown that higher CO2 levels lead to reduced plant uptake of nitrogen (and a smaller number showing the same for trace elements such as zinc) resulting in crops with lower nutritional value.'],
]
scores = model.predict(pairs)
print(scores)
# [0.8108 0.66   0.4774 0.0022 0.9549]

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'Not only is there no scientific evidence that CO2 is a pollutant, higher CO2 concentrations actually help ecosystems support more plant and animal life.',
    [
        'At very high concentrations (100 times atmospheric concentration, or greater), carbon dioxide can be toxic to animal life, so raising the concentration to 10,000 ppm (1%) or higher for several hours will eliminate pests such as whiteflies and spider mites in a greenhouse.',
        'Higher carbon dioxide concentrations will favourably affect plant growth and demand for water.',
        'Plants can grow as much as 50 percent faster in concentrations of 1,000 ppm CO 2 when compared with ambient conditions, though this assumes no change in climate and no limitation on other nutrients.',
        'Use of fertilizers are beneficial in providing nutrients to plants although they have some negative environmental effects.',
        'Studies have shown that higher CO2 levels lead to reduced plant uptake of nitrogen (and a smaller number showing the same for trace elements such as zinc) resulting in crops with lower nutritional value.',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Reranking

Dataset: claims-rerank-dev
Evaluated with CrossEncoderRerankingEvaluator with these parameters:
```
{
    "at_k": 5
}
```

Metric	Value
map	0.0653
mrr@5	0.0694
ndcg@5	0.0331

Training Details

Training Dataset

Unnamed Dataset

Size: 18,858 training samples
Columns: sentence1, sentence2, and label
Approximate statistics based on the first 1000 samples:
sentence1 sentence2 label
type string string float
details
min: 9 tokens
mean: 26.48 tokens
max: 54 tokens

min: 4 tokens
mean: 33.89 tokens
max: 475 tokens

min: 0.0
mean: 0.24
max: 1.0

	sentence1	sentence2	label
type	string	string	float
details	min: 9 tokens mean: 26.48 tokens max: 54 tokens	min: 4 tokens mean: 33.89 tokens max: 475 tokens	min: 0.0 mean: 0.24 max: 1.0

Samples:

sentence1	sentence2	label
`Not only is there no scientific evidence that CO2 is a pollutant, higher CO2 concentrations actually help ecosystems support more plant and animal life.`	`At very high concentrations (100 times atmospheric concentration, or greater), carbon dioxide can be toxic to animal life, so raising the concentration to 10,000 ppm (1%) or higher for several hours will eliminate pests such as whiteflies and spider mites in a greenhouse.`	`1.0`
`Not only is there no scientific evidence that CO2 is a pollutant, higher CO2 concentrations actually help ecosystems support more plant and animal life.`	`Higher carbon dioxide concentrations will favourably affect plant growth and demand for water.`	`1.0`
`Not only is there no scientific evidence that CO2 is a pollutant, higher CO2 concentrations actually help ecosystems support more plant and animal life.`	`Plants can grow as much as 50 percent faster in concentrations of 1,000 ppm CO 2 when compared with ambient conditions, though this assumes no change in climate and no limitation on other nutrients.`	`1.0`

Loss: BinaryCrossEntropyLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
learning_rate: 1e-06
weight_decay: 0.01
num_train_epochs: 1
warmup_steps: 0.1
fp16: True
load_best_model_at_end: True

All Hyperparameters

Click to expand

do_predict: False
prediction_loss_only: True
per_device_train_batch_size: 16
per_device_eval_batch_size: 8
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 1e-06
weight_decay: 0.01
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: None
warmup_ratio: None
warmup_steps: 0.1
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
enable_jit_checkpoint: False
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
use_cpu: False
seed: 42
data_seed: None
bf16: False
fp16: True
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: -1
ddp_backend: None
debug: []
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_prefetch_factor: None
disable_tqdm: False
remove_unused_columns: True
label_names: None
load_best_model_at_end: True
ignore_data_skip: False
fsdp: []
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
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
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
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_for_metrics: []
eval_do_concat_batches: True
auto_find_batch_size: False
full_determinism: False
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
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
use_cache: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	claims-rerank-dev_ndcg@5
0.0848	100	1.1276	0.0388
0.1696	200	1.0529	0.0371
0.2545	300	1.0724	0.0357
0.3393	400	0.9204	0.0342
0.4241	500	0.8423	0.0329
0.5089	600	0.8579	0.0351
0.5937	700	0.7598	0.0343
0.6785	800	0.7950	0.0338
0.7634	900	0.7496	0.0363
0.8482	1000	0.7760	0.0339
0.9330	1100	0.7603	0.0331

The bold row denotes the saved checkpoint.

Training Time

Training: 5.0 minutes

Framework Versions

Python: 3.12.13
Sentence Transformers: 5.4.1
Transformers: 5.0.0
PyTorch: 2.10.0+cu128
Accelerate: 1.13.0
Datasets: 4.0.0
Tokenizers: 0.22.2

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",
}