fine_tuned_model_4 / README.md

srikarvar

Add new SentenceTransformer model.

3fa6459 verified about 1 year ago

preview code

raw

history blame contribute delete

26 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:1273
  - loss:OnlineContrastiveLoss
widget:
  - source_sentence: Where can I buy organic vegetables?
    sentences:
      - Primary export product of Saudi Arabia
      - Share info about Amazon
      - Where can I buy organic fruits?
  - source_sentence: How to open a bank account in the UK?
    sentences:
      - Steps to open a bank account in the United Kingdom
      - How many weeks in a month?
      - What is the process for turning in an expense report?
  - source_sentence: What is the population of Tokyo?
    sentences:
      - What is the smallest planet in the solar system?
      - Author of the play 'Hamlet'
      - What is the population of Osaka?
  - source_sentence: How to visit the Great Wall of China?
    sentences:
      - Where can I buy a new laptop?
      - How do I close a bank account?
      - Guide to visiting the Great Wall of China
  - source_sentence: Who is the President of the United States?
    sentences:
      - What is the velocity of sound?
      - Who is the current US President?
      - History of the Byzantine Empire
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: 0.6206896551724138
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.9036016464233398
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.7192575406032483
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.9036016464233398
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.5827067669172933
            name: Cosine Precision
          - type: cosine_recall
            value: 0.9393939393939394
            name: Cosine Recall
          - type: cosine_ap
            value: 0.6366493234478966
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.6206896551724138
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 0.9036016464233398
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.7192575406032483
            name: Dot F1
          - type: dot_f1_threshold
            value: 0.9036016464233398
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.5827067669172933
            name: Dot Precision
          - type: dot_recall
            value: 0.9393939393939394
            name: Dot Recall
          - type: dot_ap
            value: 0.6366493234478966
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.6175548589341693
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 6.501791000366211
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.7232142857142857
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 7.142887115478516
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.5724381625441696
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.9818181818181818
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.64137074777591
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.6206896551724138
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 0.43908166885375977
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.7192575406032483
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 0.43908166885375977
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.5827067669172933
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.9393939393939394
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.6366493234478966
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.6206896551724138
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 6.501791000366211
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.7232142857142857
            name: Max F1
          - type: max_f1_threshold
            value: 7.142887115478516
            name: Max F1 Threshold
          - type: max_precision
            value: 0.5827067669172933
            name: Max Precision
          - type: max_recall
            value: 0.9818181818181818
            name: Max Recall
          - type: max_ap
            value: 0.64137074777591
            name: Max Ap
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: pair class test
          type: pair-class-test
        metrics:
          - type: cosine_accuracy
            value: 0.8934169278996865
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.7770164012908936
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.9034090909090907
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.7750071287155151
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.8502673796791443
            name: Cosine Precision
          - type: cosine_recall
            value: 0.9636363636363636
            name: Cosine Recall
          - type: cosine_ap
            value: 0.9467412947017336
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.8934169278996865
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 0.7770164012908936
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.9034090909090907
            name: Dot F1
          - type: dot_f1_threshold
            value: 0.7750071287155151
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.8502673796791443
            name: Dot Precision
          - type: dot_recall
            value: 0.9636363636363636
            name: Dot Recall
          - type: dot_ap
            value: 0.9467412947017336
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.890282131661442
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 9.908584594726562
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.9002849002849003
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 10.437429428100586
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.8494623655913979
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.9575757575757575
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.9451852140210413
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.8934169278996865
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 0.6678076386451721
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.9034090909090907
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 0.6708062887191772
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.8502673796791443
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.9636363636363636
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.9467412947017336
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.8934169278996865
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 9.908584594726562
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.9034090909090907
            name: Max F1
          - type: max_f1_threshold
            value: 10.437429428100586
            name: Max F1 Threshold
          - type: max_precision
            value: 0.8502673796791443
            name: Max Precision
          - type: max_recall
            value: 0.9636363636363636
            name: Max Recall
          - type: max_ap
            value: 0.9467412947017336
            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/fine_tuned_model_4")
# Run inference
sentences = [
    'Who is the President of the United States?',
    'Who is the current US President?',
    'What is the velocity of sound?',
]
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	0.6207
cosine_accuracy_threshold	0.9036
cosine_f1	0.7193
cosine_f1_threshold	0.9036
cosine_precision	0.5827
cosine_recall	0.9394
cosine_ap	0.6366
dot_accuracy	0.6207
dot_accuracy_threshold	0.9036
dot_f1	0.7193
dot_f1_threshold	0.9036
dot_precision	0.5827
dot_recall	0.9394
dot_ap	0.6366
manhattan_accuracy	0.6176
manhattan_accuracy_threshold	6.5018
manhattan_f1	0.7232
manhattan_f1_threshold	7.1429
manhattan_precision	0.5724
manhattan_recall	0.9818
manhattan_ap	0.6414
euclidean_accuracy	0.6207
euclidean_accuracy_threshold	0.4391
euclidean_f1	0.7193
euclidean_f1_threshold	0.4391
euclidean_precision	0.5827
euclidean_recall	0.9394
euclidean_ap	0.6366
max_accuracy	0.6207
max_accuracy_threshold	6.5018
max_f1	0.7232
max_f1_threshold	7.1429
max_precision	0.5827
max_recall	0.9818
max_ap	0.6414

Binary Classification

Dataset: pair-class-test
Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.8934
cosine_accuracy_threshold	0.777
cosine_f1	0.9034
cosine_f1_threshold	0.775
cosine_precision	0.8503
cosine_recall	0.9636
cosine_ap	0.9467
dot_accuracy	0.8934
dot_accuracy_threshold	0.777
dot_f1	0.9034
dot_f1_threshold	0.775
dot_precision	0.8503
dot_recall	0.9636
dot_ap	0.9467
manhattan_accuracy	0.8903
manhattan_accuracy_threshold	9.9086
manhattan_f1	0.9003
manhattan_f1_threshold	10.4374
manhattan_precision	0.8495
manhattan_recall	0.9576
manhattan_ap	0.9452
euclidean_accuracy	0.8934
euclidean_accuracy_threshold	0.6678
euclidean_f1	0.9034
euclidean_f1_threshold	0.6708
euclidean_precision	0.8503
euclidean_recall	0.9636
euclidean_ap	0.9467
max_accuracy	0.8934
max_accuracy_threshold	9.9086
max_f1	0.9034
max_f1_threshold	10.4374
max_precision	0.8503
max_recall	0.9636
max_ap	0.9467

Training Details

Training Dataset

Unnamed Dataset

Size: 1,273 training samples
Columns: sentence1, label, and sentence2
Approximate statistics based on the first 1000 samples:
sentence1 label sentence2
type string int string
details
min: 6 tokens
mean: 10.93 tokens
max: 28 tokens

0: ~48.90%
1: ~51.10%

min: 5 tokens
mean: 10.29 tokens
max: 22 tokens

	sentence1	label	sentence2
type	string	int	string
details	min: 6 tokens mean: 10.93 tokens max: 28 tokens	0: ~48.90% 1: ~51.10%	min: 5 tokens mean: 10.29 tokens max: 22 tokens

Samples:

sentence1	label	sentence2
`What are the main ingredients in a traditional pizza Margherita?`	`1`	`What ingredients are used in a classic pizza Margherita?`
`Release date of the iPhone 14`	`0`	`Release date of the iPhone 13`
`Who won the first Nobel Prize in Literature?`	`0`	`Who won the first Nobel Prize in Peace?`

Loss: OnlineContrastiveLoss

Evaluation Dataset

Unnamed Dataset

Size: 319 evaluation samples
Columns: sentence1, label, and sentence2
Approximate statistics based on the first 1000 samples:
sentence1 label sentence2
type string int string
details
min: 6 tokens
mean: 11.12 tokens
max: 22 tokens

0: ~48.28%
1: ~51.72%

min: 4 tokens
mean: 10.52 tokens
max: 21 tokens

	sentence1	label	sentence2
type	string	int	string
details	min: 6 tokens mean: 11.12 tokens max: 22 tokens	0: ~48.28% 1: ~51.72%	min: 4 tokens mean: 10.52 tokens max: 21 tokens

Samples:

sentence1	label	sentence2
`How many bones are in the human body?`	`1`	`Total bones in an adult human`
`What is the price of an iPhone 12?`	`0`	`What is the price of an iPhone 11?`
`What are the different types of renewable energy?`	`1`	`What are the various forms of renewable energy?`

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
num_train_epochs: 4
warmup_ratio: 0.1
load_best_model_at_end: True
optim: adamw_torch_fused
batch_sampler: no_duplicates

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: 5e-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: 4
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: 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: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	pair-class-dev_max_ap	pair-class-test_max_ap
0	0	-	-	0.6414	-
0.5	10	1.9407	-	-	-
1.0	20	0.9729	0.6810	-	-
1.475	30	0.4822	-	-	-
1.975	40	0.4062	-	-	-
2.025	41	-	0.5953	-	-
2.45	50	0.2894	-	-	-
2.95	60	0.1977	-	-	-
3.0	61	-	0.5318	-	-
3.425	70	0.1999	-	-	-
3.925	80	0.1491	0.5159	-	0.9467

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",
}