Redis semantic caching CrossEncoder model fine-tuned on Quora Question Pairs

This is a Cross Encoder model finetuned from cross-encoder/ms-marco-MiniLM-L12-v2 on the Quora Question Pairs LangCache Train Set dataset using the sentence-transformers library. It computes scores for pairs of texts, which can be used for sentence pair classification.

Model Details

Model Description

Model Type: Cross Encoder
Base model: cross-encoder/ms-marco-MiniLM-L12-v2
Maximum Sequence Length: 512 tokens
Number of Output Labels: 1 label
Training Dataset:
- Quora Question Pairs LangCache Train Set
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("aditeyabaral-redis/langcache-crossencoder-v1-ms-marco-MiniLM-L12-v2")
# Get scores for pairs of texts
pairs = [
    ['How can I get a list of my Gmail accounts?', 'How can I find all my old Gmail accounts?'],
    ['How can I stop Quora from modifying and editing other people’s questions on Quora?', 'Can I prevent a Quora user from editing my question on Quora?'],
    ['How much does it cost to design a logo in india?', 'How much does it cost to design a logo?'],
    ['What is screenedrenters.com?', 'What is allmyapps.com?'],
    ['What are the best colleges for an MBA in Australia?', 'What are the top MBA schools in Australia?'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'How can I get a list of my Gmail accounts?',
    [
        'How can I find all my old Gmail accounts?',
        'Can I prevent a Quora user from editing my question on Quora?',
        'How much does it cost to design a logo?',
        'What is allmyapps.com?',
        'What are the top MBA schools in Australia?',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Classification

Dataset: quora-eval
Evaluated with CrossEncoderClassificationEvaluator

Metric	Value
accuracy	0.6801
accuracy_threshold	3.2522
f1	0.5699
f1_threshold	2.8498
precision	0.4213
recall	0.8806
average_precision	0.5877

Training Details

Training Dataset

Quora Question Pairs LangCache Train Set

Dataset: Quora Question Pairs LangCache Train Set
Size: 363,861 training samples
Columns: sentence1, sentence2, and label

Approximate statistics based on the first 1000 samples:

	sentence1	sentence2	label
type	string	string	int
details	min: 15 characters mean: 60.22 characters max: 229 characters	min: 14 characters mean: 60.0 characters max: 274 characters	0: ~63.50% 1: ~36.50%

Samples:

sentence1	sentence2	label
`Why do people believe in God and how can they say he/she exists?`	`Why do we kill each other in the name of God?`	`0`
`What are the chances of a bee sting when a bee buzzes around you?`	`How can I tell if my bees are agitated/likely to sting?`	`0`
`If a man from Syro Malankara church marries a Syro-Malabar girl, can they join a Syro-Malabar parish?`	`Is Malabar Hills of Mumbai anyhow related to Malabar of Kerala?`	`0`

Loss: BinaryCrossEntropyLoss with these parameters:

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

Evaluation Dataset

Quora Question Pairs LangCache Validation Set

Dataset: Quora Question Pairs LangCache Validation Set
Size: 40,429 evaluation samples
Columns: sentence1, sentence2, and label

Approximate statistics based on the first 1000 samples:

	sentence1	sentence2	label
type	string	string	int
details	min: 13 characters mean: 59.91 characters max: 266 characters	min: 13 characters mean: 59.51 characters max: 293 characters	0: ~63.80% 1: ~36.20%

Samples:

sentence1	sentence2	label
`How can I get a list of my Gmail accounts?`	`How can I find all my old Gmail accounts?`	`1`
`How can I stop Quora from modifying and editing other people’s questions on Quora?`	`Can I prevent a Quora user from editing my question on Quora?`	`1`
`How much does it cost to design a logo in india?`	`How much does it cost to design a logo?`	`0`

Loss: BinaryCrossEntropyLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 64
per_device_eval_batch_size: 64
learning_rate: 0.0002
num_train_epochs: 15
load_best_model_at_end: True
push_to_hub: True
hub_model_id: aditeyabaral-redis/langcache-crossencoder-v1-ms-marco-MiniLM-L12-v2

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: 0.0002
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 15
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: 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}
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: True
resume_from_checkpoint: None
hub_model_id: aditeyabaral-redis/langcache-crossencoder-v1-ms-marco-MiniLM-L12-v2
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