ms-marco-MiniLM-L-6-v2 Finetuned on PV211 HomeWork

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
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("maennyn/pv211_beir_cqadupstack_crossencoder2")
# Get scores for pairs of texts
pairs = [
    ['Increase the X length of a tikzpicture', "In recent years I've developed a habit of formatting SQL `SELECT` queries like so:               SELECT         fieldNames     FROM         sources         JOIN tableSource ON col1 = col2         JOIN (             SELECT                 fieldNames             FROM                 otherSources          ) AS subQuery ON subQuery.foo = col2     WHERE         someField = somePredicate      So you see my pattern: each keyword is on its own line and that keyword's fields are indented by 1 tab-stop and the pattern is used recursively for sub- queries. This works well for all of my `SELECT` queries, as it maximizes readability though at the cost of vertical space; but it doesn't work for things like `INSERT` and `UPDATE` which have radically different syntax.               INSERT INTO tableName            (  col1,   col2,   col3,   col4,   col5,   col6,   col7,   col8  )     VALUES ( 'col1', 'col2', 'col3', 'col4', 'col5', 'col6', 'col7', 'col8' ),     VALUES ( 'col1', 'col2', 'col3', 'col4', 'col5', 'col6', 'col7', 'col8' )          UPDATE tableName     SET         col1 = 'col1',         col2 = 'col2',         col3 = 'col3',         // etc     WHERE         someField = somePredicate      As you can see, they aren't as pretty, and when you're dealing with tables with a lot of columns they quickly become unweildly. Is there a better way to format `INSERT` and `UPDATE`? And what about `CREATE` statements and other operations?"],
    ['Fillable form: checkbox linked to hide/unhide sections; pushbutton to add/delete rows', "I'd like to create a LaTeX document that when rendered into PDF, has forms that can be filled out using Adobe Reader or other such programs. Then I'd like to be able to extract the data. I deliberately would like to avoid using Acrobat for all the usual reasons (non-free, need different versions for different platforms etc). Can this be done ?"],
    ['Is there any way to get something like pmatrix with customizable grid lines between cells?', "> **Possible Duplicate:**   >  Highlight elements in the matrix i have a matrix:               \\begin{equation}      \\begin{bmatrix}         1 & 5 & 4 & 2 & 1 \\\\         1 & 5 & 4 & 2 & 1 \\\\         1 & 5 & 4 & 2 & 1 \\\\     \\end{bmatrix}     \\label{e:crop1}     \\end{equation}      and i would like to draw a box around a few of the values to highlight a selection & label it, how would i go about this? I've looked at nodes but havent got a clue. thanks"],
    ["Difference between 'all' and 'all the'", 'I am not confident about my judgement as to whether or not "the" is required if a relative clause is used in a sentence.   For example, > The data can be collected on all the computers on which the software is > installed. I think it must be "all the computers " and not be "all computers" because "computers" is specified by "on which the software is installed". Please help me confirm that I am right.'],
    ['Understanding the exclamation mark (!) in bash', "I'm following through a tutorial and it mentions to run this command:               sudo chmod 700 !$      I'm not familiar with `!$`. What does it mean?"],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'Increase the X length of a tikzpicture',
    [
        "In recent years I've developed a habit of formatting SQL `SELECT` queries like so:               SELECT         fieldNames     FROM         sources         JOIN tableSource ON col1 = col2         JOIN (             SELECT                 fieldNames             FROM                 otherSources          ) AS subQuery ON subQuery.foo = col2     WHERE         someField = somePredicate      So you see my pattern: each keyword is on its own line and that keyword's fields are indented by 1 tab-stop and the pattern is used recursively for sub- queries. This works well for all of my `SELECT` queries, as it maximizes readability though at the cost of vertical space; but it doesn't work for things like `INSERT` and `UPDATE` which have radically different syntax.               INSERT INTO tableName            (  col1,   col2,   col3,   col4,   col5,   col6,   col7,   col8  )     VALUES ( 'col1', 'col2', 'col3', 'col4', 'col5', 'col6', 'col7', 'col8' ),     VALUES ( 'col1', 'col2', 'col3', 'col4', 'col5', 'col6', 'col7', 'col8' )          UPDATE tableName     SET         col1 = 'col1',         col2 = 'col2',         col3 = 'col3',         // etc     WHERE         someField = somePredicate      As you can see, they aren't as pretty, and when you're dealing with tables with a lot of columns they quickly become unweildly. Is there a better way to format `INSERT` and `UPDATE`? And what about `CREATE` statements and other operations?",
        "I'd like to create a LaTeX document that when rendered into PDF, has forms that can be filled out using Adobe Reader or other such programs. Then I'd like to be able to extract the data. I deliberately would like to avoid using Acrobat for all the usual reasons (non-free, need different versions for different platforms etc). Can this be done ?",
        "> **Possible Duplicate:**   >  Highlight elements in the matrix i have a matrix:               \\begin{equation}      \\begin{bmatrix}         1 & 5 & 4 & 2 & 1 \\\\         1 & 5 & 4 & 2 & 1 \\\\         1 & 5 & 4 & 2 & 1 \\\\     \\end{bmatrix}     \\label{e:crop1}     \\end{equation}      and i would like to draw a box around a few of the values to highlight a selection & label it, how would i go about this? I've looked at nodes but havent got a clue. thanks",
        'I am not confident about my judgement as to whether or not "the" is required if a relative clause is used in a sentence.   For example, > The data can be collected on all the computers on which the software is > installed. I think it must be "all the computers " and not be "all computers" because "computers" is specified by "on which the software is installed". Please help me confirm that I am right.',
        "I'm following through a tutorial and it mentions to run this command:               sudo chmod 700 !$      I'm not familiar with `!$`. What does it mean?",
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Correlation

Dataset: sts_dev
Evaluated with CrossEncoderCorrelationEvaluator

Metric	Value
pearson	0.8858
spearman	0.8182

Cross Encoder Reranking

Datasets: NanoMSMARCO_R100, NanoNFCorpus_R100 and NanoNQ_R100

Evaluated with CrossEncoderRerankingEvaluator with these parameters:

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

Metric	NanoMSMARCO_R100	NanoNFCorpus_R100	NanoNQ_R100
map	0.6048 (+0.1152)	0.3633 (+0.1023)	0.6871 (+0.2674)
mrr@10	0.5974 (+0.1199)	0.5961 (+0.0962)	0.7117 (+0.2850)
ndcg@10	0.6644 (+0.1240)	0.4082 (+0.0832)	0.7413 (+0.2407)

Cross Encoder Nano BEIR

Dataset: NanoBEIR_R100_mean

Evaluated with CrossEncoderNanoBEIREvaluator with these parameters:

{
    "dataset_names": [
        "msmarco",
        "nfcorpus",
        "nq"
    ],
    "rerank_k": 100,
    "at_k": 10,
    "always_rerank_positives": true
}

Metric	Value
map	0.5517 (+0.1616)
mrr@10	0.6350 (+0.1670)
ndcg@10	0.6046 (+0.1493)

Training Details

Training Dataset

Unnamed Dataset

Size: 36,728 training samples
Columns: query, document, and label

Approximate statistics based on the first 1000 samples:

	query	document	label
type	string	string	int
details	min: 15 characters mean: 49.89 characters max: 128 characters	min: 36 characters mean: 718.8 characters max: 17541 characters	0: ~48.90% 1: ~51.10%

Samples:

query	document	label
`Increase the X length of a tikzpicture`	In recent years I've developed a habit of formatting SQL SELECT queries like so: SELECT fieldNames FROM sources JOIN tableSource ON col1 = col2 JOIN ( SELECT fieldNames FROM otherSources ) AS subQuery ON subQuery.foo = col2 WHERE someField = somePredicate So you see my pattern: each keyword is on its own line and that keyword's fields are indented by 1 tab-stop and the pattern is used recursively for sub- queries. This works well for all of my SELECT queries, as it maximizes readability though at the cost of vertical space; but it doesn't work for things like INSERT and UPDATE which have radically different syntax. INSERT INTO tableName ( col1, col2, col3, col4, col5, col6, col7, col8 ) VALUES ( 'col1', 'col2', 'col3', 'col4', 'col5', 'col6', 'col7', 'col8' ), VALUES ( 'col1', 'col2', 'col3', 'col4',...	`0`
`Fillable form: checkbox linked to hide/unhide sections; pushbutton to add/delete rows`	`I'd like to create a LaTeX document that when rendered into PDF, has forms that can be filled out using Adobe Reader or other such programs. Then I'd like to be able to extract the data. I deliberately would like to avoid using Acrobat for all the usual reasons (non-free, need different versions for different platforms etc). Can this be done ?`	`1`
`Is there any way to get something like pmatrix with customizable grid lines between cells?`	`> Possible Duplicate: > Highlight elements in the matrix i have a matrix: \begin{equation} \begin{bmatrix} 1 & 5 & 4 & 2 & 1 \ 1 & 5 & 4 & 2 & 1 \ 1 & 5 & 4 & 2 & 1 \ \end{bmatrix} \label{e:crop1} \end{equation} and i would like to draw a box around a few of the values to highlight a selection & label it, how would i go about this? I've looked at nodes but havent got a clue. thanks`	`1`

Loss: BinaryCrossEntropyLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
learning_rate: 2e-05
warmup_ratio: 0.1
save_only_model: True
fp16: True
load_best_model_at_end: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
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: 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: 3
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: True
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: True
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}
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	sts_dev_spearman	NanoMSMARCO_R100_ndcg@10	NanoNFCorpus_R100_ndcg@10	NanoNQ_R100_ndcg@10	NanoBEIR_R100_mean_ndcg@10
-1	-1	-	0.7222	0.6686 (+0.1282)	0.3930 (+0.0680)	0.7599 (+0.2592)	0.6072 (+0.1518)
0.4355	1000	0.4163	-	-	-	-	-
0.8711	2000	0.1632	-	-	-	-	-
1.0	2296	-	0.8182	0.6644 (+0.1240)	0.4082 (+0.0832)	0.7413 (+0.2407)	0.6046 (+0.1493)
1.3066	3000	0.1227	-	-	-	-	-
1.7422	4000	0.1157	-	-	-	-	-
2.0	4592	-	0.8201	0.6266 (+0.0862)	0.4096 (+0.0846)	0.7032 (+0.2026)	0.5798 (+0.1244)
2.1777	5000	0.0964	-	-	-	-	-
2.6132	6000	0.081	-	-	-	-	-
3.0	6888	-	0.8203	0.6241 (+0.0837)	0.4068 (+0.0817)	0.6931 (+0.1924)	0.5747 (+0.1193)
-1	-1	-	0.8182	0.6644 (+0.1240)	0.4082 (+0.0832)	0.7413 (+0.2407)	0.6046 (+0.1493)

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.11.11
Sentence Transformers: 4.1.0
Transformers: 4.51.3
PyTorch: 2.8.0.dev20250319+cu128
Accelerate: 1.6.0
Datasets: 3.6.0
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",
}