SentenceTransformer based on Alibaba-NLP/gte-modernbert-base

This is a sentence-transformers model finetuned from Alibaba-NLP/gte-modernbert-base on the csv dataset. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: Alibaba-NLP/gte-modernbert-base
Maximum Sequence Length: 8192 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- csv

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: ModernBertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

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 SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("waris-gill/ModernBert-Quora-V1")
# Run inference
sentences = [
    'Will the value of Indian rupee increase after the ban of 500 and 1000 rupee notes?',
    'What will be the implications of banning 500 and 1000 rupees currency notes on Indian economy?',
    "Are Danish Sait's prank calls fake?",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

My Binary Classification

Evaluated with scache.train.MyBinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.9023
cosine_accuracy_threshold	0.8583
cosine_f1	0.9028
cosine_f1_threshold	0.8583
cosine_precision	0.8987
cosine_recall	0.9069
cosine_ap	0.952
cosine_mcc	0.8047

Training Details

Training Dataset

csv

Dataset: csv
Size: 323,480 training samples
Columns: question_1, question_2, and label
Approximate statistics based on the first 1000 samples:
question_1 question_2 label
type string string int
details
min: 6 tokens
mean: 15.26 tokens
max: 69 tokens

min: 3 tokens
mean: 15.45 tokens
max: 64 tokens

0: ~61.90%
1: ~38.10%

	question_1	question_2	label
type	string	string	int
details	min: 6 tokens mean: 15.26 tokens max: 69 tokens	min: 3 tokens mean: 15.45 tokens max: 64 tokens	0: ~61.90% 1: ~38.10%

Samples:

question_1	question_2	label
`What is mutual funds with examples?`	`What is mutual fund all about?`	`1`
`What are holograms?`	`What are the possible applications of holographic imaging?`	`0`
`What is effective mass?`	`What is the effective mass of silicon?`	`0`

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "OnlineContrastiveLoss",
    "matryoshka_dims": [
        768,
        512,
        256,
        128,
        64
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Evaluation Dataset

csv

Dataset: csv
Size: 323,480 evaluation samples
Columns: question_1, question_2, and label
Approximate statistics based on the first 1000 samples:
question_1 question_2 label
type string string int
details
min: 6 tokens
mean: 15.45 tokens
max: 58 tokens

min: 7 tokens
mean: 15.44 tokens
max: 72 tokens

0: ~50.30%
1: ~49.70%

	question_1	question_2	label
type	string	string	int
details	min: 6 tokens mean: 15.45 tokens max: 58 tokens	min: 7 tokens mean: 15.44 tokens max: 72 tokens	0: ~50.30% 1: ~49.70%

Samples:

question_1	question_2	label
`How does one stop taking oneself too seriously?`	`How do I stop taking myself so seriously?`	`1`
`Why do English people like football?`	`Why do English people love football so much?`	`1`
`What is your analysis of the US Presidential political debate 9/26/2016?`	`What do you think of the first US presidential debate?`	`1`

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "OnlineContrastiveLoss",
    "matryoshka_dims": [
        768,
        512,
        256,
        128,
        64
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
per_device_eval_batch_size: 256
learning_rate: 6.5383156211679e-05
max_grad_norm: 0.5
num_train_epochs: 1
lr_scheduler_type: constant
load_best_model_at_end: True
torch_compile: True
torch_compile_backend: inductor
batch_sampler: no_duplicates

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: 256
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: 6.5383156211679e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 0.5
num_train_epochs: 1
max_steps: -1
lr_scheduler_type: constant
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: 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: True
torch_compile_backend: inductor
torch_compile_mode: None
dispatch_batches: None
split_batches: 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: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs