fine_tuned_model_11 / README.md

srikarvar

Add new SentenceTransformer model.

be17ba2 verified over 1 year ago

preview code

raw

history blame contribute delete

26.6 kB

metadata

base_model: intfloat/multilingual-e5-small
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:1936
  - loss:OnlineContrastiveLoss
widget:
  - source_sentence: What are the symptoms of COVID-19?
    sentences:
      - How to identify COVID-19?
      - What is the process for booking a dinner table?
      - >-
        It is not necessary to include specific fields in a financial report;
        nevertheless, it is beneficial to add pertinent financial metrics to
        help investors gauge the company's condition.
  - source_sentence: How to apply for a scholarship?
    sentences:
      - Steps to apply for a scholarship
      - Advantages of practicing meditation
      - >-
        When `ignore_metadata` is set to `True`, all metadata and attributes are
        stripped from the file prior to processing.
  - source_sentence: How to write a novel?
    sentences:
      - How to write a short story?
      - Who wrote 'Macbeth'?
      - How to reset a phone
  - source_sentence: >-
      You can wrap the project in `job.utils.data.JobLoader` and create a
      collate function to collate the tasks into batches.
    sentences:
      - Steps to prepare a steak
      - How many people live in Germany?
      - >-
        You can use `job.utils.data.JobLoader` to encapsulate the project and
        define a collate function to group the tasks into batches.
  - source_sentence: What is the time now?
    sentences:
      - How to cook a chicken?
      - Current time
      - Guide to starting a small business
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: 0.9212962962962963
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.8385236263275146
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.9403508771929825
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.8385236263275146
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.9370629370629371
            name: Cosine Precision
          - type: cosine_recall
            value: 0.9436619718309859
            name: Cosine Recall
          - type: cosine_ap
            value: 0.9872231100578164
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.9212962962962963
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 0.8385236263275146
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.9403508771929825
            name: Dot F1
          - type: dot_f1_threshold
            value: 0.8385236263275146
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.9370629370629371
            name: Dot Precision
          - type: dot_recall
            value: 0.9436619718309859
            name: Dot Recall
          - type: dot_ap
            value: 0.9872231100578164
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.9166666666666666
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 8.658426284790039
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.9391891891891893
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 9.594137191772461
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.9025974025974026
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.9788732394366197
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.987218816132896
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.9212962962962963
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 0.568278431892395
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.9403508771929825
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 0.568278431892395
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.9370629370629371
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.9436619718309859
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.9872231100578164
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.9212962962962963
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 8.658426284790039
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.9403508771929825
            name: Max F1
          - type: max_f1_threshold
            value: 9.594137191772461
            name: Max F1 Threshold
          - type: max_precision
            value: 0.9370629370629371
            name: Max Precision
          - type: max_recall
            value: 0.9788732394366197
            name: Max Recall
          - type: max_ap
            value: 0.9872231100578164
            name: Max Ap
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: pair class test
          type: pair-class-test
        metrics:
          - type: cosine_accuracy
            value: 0.9305555555555556
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.8569861650466919
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.9484536082474226
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.8531842827796936
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.9261744966442953
            name: Cosine Precision
          - type: cosine_recall
            value: 0.971830985915493
            name: Cosine Recall
          - type: cosine_ap
            value: 0.9898045699188958
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.9305555555555556
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 0.8569861650466919
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.9484536082474226
            name: Dot F1
          - type: dot_f1_threshold
            value: 0.8531842231750488
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.9261744966442953
            name: Dot Precision
          - type: dot_recall
            value: 0.971830985915493
            name: Dot Recall
          - type: dot_ap
            value: 0.9898045699188958
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.9351851851851852
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 8.299823760986328
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.9517241379310345
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 8.299823760986328
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.9324324324324325
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.971830985915493
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.9895380844501982
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.9305555555555556
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 0.534814715385437
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.9484536082474226
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 0.5418605804443359
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.9261744966442953
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.971830985915493
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.9898045699188958
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.9351851851851852
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 8.299823760986328
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.9517241379310345
            name: Max F1
          - type: max_f1_threshold
            value: 8.299823760986328
            name: Max F1 Threshold
          - type: max_precision
            value: 0.9324324324324325
            name: Max Precision
          - type: max_recall
            value: 0.971830985915493
            name: Max Recall
          - type: max_ap
            value: 0.9898045699188958
            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_11")
# Run inference
sentences = [
    'What is the time now?',
    'Current time',
    'Guide to starting a small business',
]
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	0.9213
cosine_accuracy_threshold	0.8385
cosine_f1	0.9404
cosine_f1_threshold	0.8385
cosine_precision	0.9371
cosine_recall	0.9437
cosine_ap	0.9872
dot_accuracy	0.9213
dot_accuracy_threshold	0.8385
dot_f1	0.9404
dot_f1_threshold	0.8385
dot_precision	0.9371
dot_recall	0.9437
dot_ap	0.9872
manhattan_accuracy	0.9167
manhattan_accuracy_threshold	8.6584
manhattan_f1	0.9392
manhattan_f1_threshold	9.5941
manhattan_precision	0.9026
manhattan_recall	0.9789
manhattan_ap	0.9872
euclidean_accuracy	0.9213
euclidean_accuracy_threshold	0.5683
euclidean_f1	0.9404
euclidean_f1_threshold	0.5683
euclidean_precision	0.9371
euclidean_recall	0.9437
euclidean_ap	0.9872
max_accuracy	0.9213
max_accuracy_threshold	8.6584
max_f1	0.9404
max_f1_threshold	9.5941
max_precision	0.9371
max_recall	0.9789
max_ap	0.9872

Binary Classification

Dataset: pair-class-test
Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.9306
cosine_accuracy_threshold	0.857
cosine_f1	0.9485
cosine_f1_threshold	0.8532
cosine_precision	0.9262
cosine_recall	0.9718
cosine_ap	0.9898
dot_accuracy	0.9306
dot_accuracy_threshold	0.857
dot_f1	0.9485
dot_f1_threshold	0.8532
dot_precision	0.9262
dot_recall	0.9718
dot_ap	0.9898
manhattan_accuracy	0.9352
manhattan_accuracy_threshold	8.2998
manhattan_f1	0.9517
manhattan_f1_threshold	8.2998
manhattan_precision	0.9324
manhattan_recall	0.9718
manhattan_ap	0.9895
euclidean_accuracy	0.9306
euclidean_accuracy_threshold	0.5348
euclidean_f1	0.9485
euclidean_f1_threshold	0.5419
euclidean_precision	0.9262
euclidean_recall	0.9718
euclidean_ap	0.9898
max_accuracy	0.9352
max_accuracy_threshold	8.2998
max_f1	0.9517
max_f1_threshold	8.2998
max_precision	0.9324
max_recall	0.9718
max_ap	0.9898

Training Details

Training Dataset

Unnamed Dataset

Size: 1,936 training samples
Columns: label, sentence1, and sentence2
Approximate statistics based on the first 1000 samples:
label sentence1 sentence2
type int string string
details
0: ~35.30%
1: ~64.70%

min: 6 tokens
mean: 16.19 tokens
max: 98 tokens

min: 4 tokens
mean: 15.75 tokens
max: 98 tokens

	label	sentence1	sentence2
type	int	string	string
details	0: ~35.30% 1: ~64.70%	min: 6 tokens mean: 16.19 tokens max: 98 tokens	min: 4 tokens mean: 15.75 tokens max: 98 tokens

Samples:

label	sentence1	sentence2
`1`	`How do I apply for a credit card?`	`How do I get a credit card?`
`1`	`What is the function of a learning rate scheduler?`	`How does a learning rate scheduler optimize training?`
`0`	`What is the speed of a rocket?`	`What is the speed of a jet plane?`

Loss: OnlineContrastiveLoss

Evaluation Dataset

Unnamed Dataset

Size: 216 evaluation samples
Columns: label, sentence1, and sentence2
Approximate statistics based on the first 216 samples:
label sentence1 sentence2
type int string string
details
0: ~34.26%
1: ~65.74%

min: 6 tokens
mean: 15.87 tokens
max: 87 tokens

min: 4 tokens
mean: 15.61 tokens
max: 86 tokens

	label	sentence1	sentence2
type	int	string	string
details	0: ~34.26% 1: ~65.74%	min: 6 tokens mean: 15.87 tokens max: 87 tokens	min: 4 tokens mean: 15.61 tokens max: 86 tokens

Samples:

label	sentence1	sentence2
`0`	`What is the freezing point of ethanol?`	`What is the boiling point of ethanol?`
`0`	`Healthy habits`	`Unhealthy habits`
`0`	`What is the difference between omnivores and herbivores?`	`What is the difference between omnivores, carnivores, and herbivores?`

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	-	-	0.8705	-
0.3279	10	1.3831	-	-	-
0.6557	20	0.749	-	-	-
0.9836	30	0.5578	0.2991	0.9862	-
1.3115	40	0.3577	-	-	-
1.6393	50	0.2594	-	-	-
1.9672	60	0.2119	-	-	-
2.0	61	-	0.2753	0.9898	-
2.2951	70	0.17	-	-	-
2.6230	80	0.1126	-	-	-
2.9508	90	0.0538	-	-	-
2.9836	91	-	0.3222	0.9864	-
3.2787	100	0.1423	-	-	-
3.6066	110	0.066	-	-	-
3.9344	120	0.0486	0.3237	0.9872	0.9898

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.1.0
Transformers: 4.41.2
PyTorch: 2.1.2+cu121
Accelerate: 0.34.2
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",
}