multilingual-e5-small-pairclass-4 / README.md

srikarvar

Add new SentenceTransformer model.

3ba5aa4 verified about 1 year ago

preview code

raw

history blame contribute delete

28.6 kB

metadata

base_model: intfloat/multilingual-e5-small
datasets: []
language: []
library_name: sentence-transformers
metrics:
  - cosine_accuracy
  - cosine_accuracy_threshold
  - cosine_f1
  - cosine_f1_threshold
  - cosine_precision
  - cosine_recall
  - cosine_ap
  - dot_accuracy
  - dot_accuracy_threshold
  - dot_f1
  - dot_f1_threshold
  - dot_precision
  - dot_recall
  - dot_ap
  - manhattan_accuracy
  - manhattan_accuracy_threshold
  - manhattan_f1
  - manhattan_f1_threshold
  - manhattan_precision
  - manhattan_recall
  - manhattan_ap
  - euclidean_accuracy
  - euclidean_accuracy_threshold
  - euclidean_f1
  - euclidean_f1_threshold
  - euclidean_precision
  - euclidean_recall
  - euclidean_ap
  - max_accuracy
  - max_accuracy_threshold
  - max_f1
  - max_f1_threshold
  - max_precision
  - max_recall
  - max_ap
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:971
  - loss:OnlineContrastiveLoss
widget:
  - source_sentence: Steps to bake a pie
    sentences:
      - How to bake a pie?
      - What are the ingredients of a pizza?
      - How to create a business plan?
  - source_sentence: What are the benefits of yoga?
    sentences:
      - If I combine the yellow and blue colors, what color will I get?
      - Can you help me understand this contract?
      - What are the benefits of meditation?
  - source_sentence: Capital city of Canada
    sentences:
      - What time does the movie start?
      - Who is the President of the United States?
      - What is the capital of Canada?
  - source_sentence: Tell me about Shopify
    sentences:
      - Who discovered penicillin?
      - Share info about Shopify
      - Who invented the telephone?
  - source_sentence: What is the melting point of ice at sea level?
    sentences:
      - What is the boiling point of water at sea level?
      - Can you recommend a good restaurant nearby?
      - Tell me a joke
model-index:
  - name: SentenceTransformer based on intfloat/multilingual-e5-small
    results:
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: pair class dev
          type: pair-class-dev
        metrics:
          - type: cosine_accuracy
            value: 0.9300411522633745
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.788658857345581
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.9237668161434978
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.7819762825965881
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.8956521739130435
            name: Cosine Precision
          - type: cosine_recall
            value: 0.9537037037037037
            name: Cosine Recall
          - type: cosine_ap
            value: 0.9603135110633257
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.9300411522633745
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 0.788658857345581
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.9237668161434978
            name: Dot F1
          - type: dot_f1_threshold
            value: 0.7819762229919434
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.8956521739130435
            name: Dot Precision
          - type: dot_recall
            value: 0.9537037037037037
            name: Dot Recall
          - type: dot_ap
            value: 0.9603135110633257
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.9218106995884774
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 9.936657905578613
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.914798206278027
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 10.316186904907227
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.8869565217391304
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.9444444444444444
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.9578931449470002
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.9300411522633745
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 0.6501401662826538
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.9237668161434978
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 0.6603381633758545
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.8956521739130435
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.9537037037037037
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.9603135110633257
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.9300411522633745
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 9.936657905578613
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.9237668161434978
            name: Max F1
          - type: max_f1_threshold
            value: 10.316186904907227
            name: Max F1 Threshold
          - type: max_precision
            value: 0.8956521739130435
            name: Max Precision
          - type: max_recall
            value: 0.9537037037037037
            name: Max Recall
          - type: max_ap
            value: 0.9603135110633257
            name: Max Ap
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: pair class test
          type: pair-class-test
        metrics:
          - type: cosine_accuracy
            value: 0.9300411522633745
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.788658857345581
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.9237668161434978
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.7819762825965881
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.8956521739130435
            name: Cosine Precision
          - type: cosine_recall
            value: 0.9537037037037037
            name: Cosine Recall
          - type: cosine_ap
            value: 0.9603135110633257
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.9300411522633745
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 0.788658857345581
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.9237668161434978
            name: Dot F1
          - type: dot_f1_threshold
            value: 0.7819762229919434
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.8956521739130435
            name: Dot Precision
          - type: dot_recall
            value: 0.9537037037037037
            name: Dot Recall
          - type: dot_ap
            value: 0.9603135110633257
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.9218106995884774
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 9.936657905578613
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.914798206278027
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 10.316186904907227
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.8869565217391304
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.9444444444444444
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.9578931449470002
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.9300411522633745
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 0.6501401662826538
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.9237668161434978
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 0.6603381633758545
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.8956521739130435
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.9537037037037037
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.9603135110633257
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.9300411522633745
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 9.936657905578613
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.9237668161434978
            name: Max F1
          - type: max_f1_threshold
            value: 10.316186904907227
            name: Max F1 Threshold
          - type: max_precision
            value: 0.8956521739130435
            name: Max Precision
          - type: max_recall
            value: 0.9537037037037037
            name: Max Recall
          - type: max_ap
            value: 0.9603135110633257
            name: Max Ap

SentenceTransformer based on intfloat/multilingual-e5-small

This is a sentence-transformers model finetuned from intfloat/multilingual-e5-small. It maps sentences & paragraphs to a 384-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: intfloat/multilingual-e5-small
Maximum Sequence Length: 512 tokens
Output Dimensionality: 384 tokens
Similarity Function: Cosine Similarity

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': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

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("srikarvar/multilingual-e5-small-pairclass-4")
# Run inference
sentences = [
    'What is the melting point of ice at sea level?',
    'What is the boiling point of water at sea level?',
    'Can you recommend a good restaurant nearby?',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

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

Evaluation

Metrics

Binary Classification

Dataset: pair-class-dev
Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.93
cosine_accuracy_threshold	0.7887
cosine_f1	0.9238
cosine_f1_threshold	0.782
cosine_precision	0.8957
cosine_recall	0.9537
cosine_ap	0.9603
dot_accuracy	0.93
dot_accuracy_threshold	0.7887
dot_f1	0.9238
dot_f1_threshold	0.782
dot_precision	0.8957
dot_recall	0.9537
dot_ap	0.9603
manhattan_accuracy	0.9218
manhattan_accuracy_threshold	9.9367
manhattan_f1	0.9148
manhattan_f1_threshold	10.3162
manhattan_precision	0.887
manhattan_recall	0.9444
manhattan_ap	0.9579
euclidean_accuracy	0.93
euclidean_accuracy_threshold	0.6501
euclidean_f1	0.9238
euclidean_f1_threshold	0.6603
euclidean_precision	0.8957
euclidean_recall	0.9537
euclidean_ap	0.9603
max_accuracy	0.93
max_accuracy_threshold	9.9367
max_f1	0.9238
max_f1_threshold	10.3162
max_precision	0.8957
max_recall	0.9537
max_ap	0.9603

Binary Classification

Dataset: pair-class-test
Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.93
cosine_accuracy_threshold	0.7887
cosine_f1	0.9238
cosine_f1_threshold	0.782
cosine_precision	0.8957
cosine_recall	0.9537
cosine_ap	0.9603
dot_accuracy	0.93
dot_accuracy_threshold	0.7887
dot_f1	0.9238
dot_f1_threshold	0.782
dot_precision	0.8957
dot_recall	0.9537
dot_ap	0.9603
manhattan_accuracy	0.9218
manhattan_accuracy_threshold	9.9367
manhattan_f1	0.9148
manhattan_f1_threshold	10.3162
manhattan_precision	0.887
manhattan_recall	0.9444
manhattan_ap	0.9579
euclidean_accuracy	0.93
euclidean_accuracy_threshold	0.6501
euclidean_f1	0.9238
euclidean_f1_threshold	0.6603
euclidean_precision	0.8957
euclidean_recall	0.9537
euclidean_ap	0.9603
max_accuracy	0.93
max_accuracy_threshold	9.9367
max_f1	0.9238
max_f1_threshold	10.3162
max_precision	0.8957
max_recall	0.9537
max_ap	0.9603

Training Details

Training Dataset

Unnamed Dataset

Size: 971 training samples
Columns: sentence2, sentence1, and label
Approximate statistics based on the first 1000 samples:
sentence2 sentence1 label
type string string int
details
min: 4 tokens
mean: 10.12 tokens
max: 22 tokens

min: 6 tokens
mean: 10.82 tokens
max: 22 tokens

0: ~48.61%
1: ~51.39%

	sentence2	sentence1	label
type	string	string	int
details	min: 4 tokens mean: 10.12 tokens max: 22 tokens	min: 6 tokens mean: 10.82 tokens max: 22 tokens	0: ~48.61% 1: ~51.39%

Samples:

sentence2	sentence1	label
`Total number of bones in an adult human body`	`How many bones are in the human body?`	`1`
`What is the largest river in North America?`	`What is the largest lake in North America?`	`0`
`What is the capital of Australia?`	`What is the capital of New Zealand?`	`0`

Loss: OnlineContrastiveLoss

Evaluation Dataset

Unnamed Dataset

Size: 243 evaluation samples
Columns: sentence2, sentence1, and label
Approximate statistics based on the first 1000 samples:
sentence2 sentence1 label
type string string int
details
min: 4 tokens
mean: 10.09 tokens
max: 20 tokens

min: 6 tokens
mean: 10.55 tokens
max: 22 tokens

0: ~55.56%
1: ~44.44%

	sentence2	sentence1	label
type	string	string	int
details	min: 4 tokens mean: 10.09 tokens max: 20 tokens	min: 6 tokens mean: 10.55 tokens max: 22 tokens	0: ~55.56% 1: ~44.44%

Samples:

sentence2	sentence1	label
`What are the various forms of renewable energy?`	`What are the different types of renewable energy?`	`1`
`Gravity discoverer`	`Who discovered gravity?`	`1`
`Can you help me write this report?`	`Can you help me understand this report?`	`0`

Loss: OnlineContrastiveLoss

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
gradient_accumulation_steps: 2
learning_rate: 3e-06
weight_decay: 0.01
num_train_epochs: 15
lr_scheduler_type: reduce_lr_on_plateau
warmup_ratio: 0.1
load_best_model_at_end: True
optim: adamw_torch_fused

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
prediction_loss_only: True
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 2
eval_accumulation_steps: None
learning_rate: 3e-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: 15
max_steps: -1
lr_scheduler_type: reduce_lr_on_plateau
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: 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_fused
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: False
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
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
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
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	pair-class-dev_max_ap	pair-class-test_max_ap
0	0	-	-	0.6426	-
0.6452	10	4.7075	-	-	-
0.9677	15	-	3.1481	0.7843	-
1.2903	20	3.431	-	-	-
1.9355	30	3.4054	-	-	-
2.0	31	-	2.1820	0.8692	-
2.5806	40	2.2735	-	-	-
2.9677	46	-	1.8185	0.9078	-
3.2258	50	2.3159	-	-	-
3.8710	60	2.1466	-	-	-
4.0	62	-	1.5769	0.9252	-
4.5161	70	1.6873	-	-	-
4.9677	77	-	1.4342	0.9310	-
5.1613	80	1.5927	-	-	-
5.8065	90	1.4184	-	-	-
6.0	93	-	1.3544	0.9357	-
6.4516	100	1.333	-	-	-
6.9677	108	-	1.2630	0.9402	-
7.0968	110	1.089	-	-	-
7.7419	120	1.0947	-	-	-
8.0	124	-	1.2120	0.9444	-
8.3871	130	0.8118	-	-	-
8.9677	139	-	1.1641	0.9454	-
9.0323	140	1.0237	-	-	-
9.6774	150	0.8406	-	-	-
10.0	155	-	1.0481	0.9464	-
10.3226	160	0.7081	-	-	-
10.9677	170	0.7397	0.9324	0.9509	-
11.6129	180	0.5604	-	-	-
12.0	186	-	0.8386	0.9556	-
12.2581	190	0.5841	-	-	-
12.9032	200	0.5463	-	-	-
12.9677	201	-	0.7930	0.9577	-
13.5484	210	0.4599	-	-	-
14.0	217	-	0.7564	0.9599	-
14.1935	220	0.2437	-	-	-
14.5161	225	-	0.7522	0.9603	0.9603

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.0.1
Transformers: 4.41.2
PyTorch: 2.1.2+cu121
Accelerate: 0.32.1
Datasets: 2.19.1
Tokenizers: 0.19.1

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",
}