fine_tuned_model_14 / README.md

srikarvar

Add new SentenceTransformer model.

738266b verified over 1 year ago

preview code

raw

history blame contribute delete

30.2 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:2836
  - loss:OnlineContrastiveLoss
widget:
  - source_sentence: No, it doesn't exist in version 5.3.1.
    sentences:
      - >-
        The `from_dictionary` function requires the following:

        - `data` (Union[dict, Mapping]): A collection of keys linked to values
        or Python objects.

        - `schema` (Schema, optional): If not given, it will be determined from
        the Mapping values.

        - `metadata` (Union[dict, Mapping], optional): Optional metadata for the
        schema (if inferred).
      - Stages of photosynthesis
      - Version 5.3.1 does not contain it.
  - source_sentence: How to make homemade ice cream?
    sentences:
      - Recipe for making ice cream at home
      - >-
        How will abolishing Rs. 500 and Rs. 1000 notes affect the real estate
        businesses in India?
      - How many people live in Japan?
  - source_sentence: Best books on World War II
    sentences:
      - How do I go about getting a visa?
      - What steps are involved in performing market analysis?
      - Top literature about World War II
  - source_sentence: What is the benefit of going Walking every morning?
    sentences:
      - What are the top workouts for losing weight?
      - How large is Japan?
      - >-
        Bollywood industry doesn't encourage outsiders? For ex outsiders may get
        one or at max two chances whereas star kids get multiple chances to
        perform?
  - source_sentence: >-
      The purpose of the training guide is to provide tutorials, how-to guides,
      and conceptual guides for working with AI models.
    sentences:
      - Steps to roast a turkey
      - >-
        The goal of the training guide is to offer tutorials, how-to
        instructions, and conceptual guidance for utilizing AI models.
      - Who was the first person to fly across the Atlantic?
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: 0.8639240506329114
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.8522839546203613
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.8853333333333334
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.8417313098907471
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.9021739130434783
            name: Cosine Precision
          - type: cosine_recall
            value: 0.8691099476439791
            name: Cosine Recall
          - type: cosine_ap
            value: 0.9514746651949948
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.8639240506329114
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 0.8522839546203613
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.8853333333333334
            name: Dot F1
          - type: dot_f1_threshold
            value: 0.8417313098907471
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.9021739130434783
            name: Dot Precision
          - type: dot_recall
            value: 0.8691099476439791
            name: Dot Recall
          - type: dot_ap
            value: 0.9514746651949948
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.8670886075949367
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 8.227925300598145
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.8877005347593583
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 8.646421432495117
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.907103825136612
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.8691099476439791
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.9520439027006086
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.8639240506329114
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 0.5435356497764587
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.8853333333333334
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 0.5626147985458374
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.9021739130434783
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.8691099476439791
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.9514724841898053
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.8670886075949367
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 8.227925300598145
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.8877005347593583
            name: Max F1
          - type: max_f1_threshold
            value: 8.646421432495117
            name: Max F1 Threshold
          - type: max_precision
            value: 0.907103825136612
            name: Max Precision
          - type: max_recall
            value: 0.8691099476439791
            name: Max Recall
          - type: max_ap
            value: 0.9520439027006086
            name: Max Ap
      - task:
          type: binary-classification
          name: Binary Classification
        dataset:
          name: pair class test
          type: pair-class-test
        metrics:
          - type: cosine_accuracy
            value: 0.870253164556962
            name: Cosine Accuracy
          - type: cosine_accuracy_threshold
            value: 0.8251076936721802
            name: Cosine Accuracy Threshold
          - type: cosine_f1
            value: 0.8935064935064936
            name: Cosine F1
          - type: cosine_f1_threshold
            value: 0.8084052801132202
            name: Cosine F1 Threshold
          - type: cosine_precision
            value: 0.8865979381443299
            name: Cosine Precision
          - type: cosine_recall
            value: 0.900523560209424
            name: Cosine Recall
          - type: cosine_ap
            value: 0.9546600352559002
            name: Cosine Ap
          - type: dot_accuracy
            value: 0.870253164556962
            name: Dot Accuracy
          - type: dot_accuracy_threshold
            value: 0.8251076936721802
            name: Dot Accuracy Threshold
          - type: dot_f1
            value: 0.8935064935064936
            name: Dot F1
          - type: dot_f1_threshold
            value: 0.808405339717865
            name: Dot F1 Threshold
          - type: dot_precision
            value: 0.8865979381443299
            name: Dot Precision
          - type: dot_recall
            value: 0.900523560209424
            name: Dot Recall
          - type: dot_ap
            value: 0.9546600352559002
            name: Dot Ap
          - type: manhattan_accuracy
            value: 0.870253164556962
            name: Manhattan Accuracy
          - type: manhattan_accuracy_threshold
            value: 9.181171417236328
            name: Manhattan Accuracy Threshold
          - type: manhattan_f1
            value: 0.8912466843501327
            name: Manhattan F1
          - type: manhattan_f1_threshold
            value: 9.181171417236328
            name: Manhattan F1 Threshold
          - type: manhattan_precision
            value: 0.9032258064516129
            name: Manhattan Precision
          - type: manhattan_recall
            value: 0.8795811518324608
            name: Manhattan Recall
          - type: manhattan_ap
            value: 0.9546014712222561
            name: Manhattan Ap
          - type: euclidean_accuracy
            value: 0.870253164556962
            name: Euclidean Accuracy
          - type: euclidean_accuracy_threshold
            value: 0.591425895690918
            name: Euclidean Accuracy Threshold
          - type: euclidean_f1
            value: 0.8935064935064936
            name: Euclidean F1
          - type: euclidean_f1_threshold
            value: 0.6190224885940552
            name: Euclidean F1 Threshold
          - type: euclidean_precision
            value: 0.8865979381443299
            name: Euclidean Precision
          - type: euclidean_recall
            value: 0.900523560209424
            name: Euclidean Recall
          - type: euclidean_ap
            value: 0.9546600352559002
            name: Euclidean Ap
          - type: max_accuracy
            value: 0.870253164556962
            name: Max Accuracy
          - type: max_accuracy_threshold
            value: 9.181171417236328
            name: Max Accuracy Threshold
          - type: max_f1
            value: 0.8935064935064936
            name: Max F1
          - type: max_f1_threshold
            value: 9.181171417236328
            name: Max F1 Threshold
          - type: max_precision
            value: 0.9032258064516129
            name: Max Precision
          - type: max_recall
            value: 0.900523560209424
            name: Max Recall
          - type: max_ap
            value: 0.9546600352559002
            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_14")
# Run inference
sentences = [
    'The purpose of the training guide is to provide tutorials, how-to guides, and conceptual guides for working with AI models.',
    'The goal of the training guide is to offer tutorials, how-to instructions, and conceptual guidance for utilizing AI models.',
    'Steps to roast a turkey',
]
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	0.8639
cosine_accuracy_threshold	0.8523
cosine_f1	0.8853
cosine_f1_threshold	0.8417
cosine_precision	0.9022
cosine_recall	0.8691
cosine_ap	0.9515
dot_accuracy	0.8639
dot_accuracy_threshold	0.8523
dot_f1	0.8853
dot_f1_threshold	0.8417
dot_precision	0.9022
dot_recall	0.8691
dot_ap	0.9515
manhattan_accuracy	0.8671
manhattan_accuracy_threshold	8.2279
manhattan_f1	0.8877
manhattan_f1_threshold	8.6464
manhattan_precision	0.9071
manhattan_recall	0.8691
manhattan_ap	0.952
euclidean_accuracy	0.8639
euclidean_accuracy_threshold	0.5435
euclidean_f1	0.8853
euclidean_f1_threshold	0.5626
euclidean_precision	0.9022
euclidean_recall	0.8691
euclidean_ap	0.9515
max_accuracy	0.8671
max_accuracy_threshold	8.2279
max_f1	0.8877
max_f1_threshold	8.6464
max_precision	0.9071
max_recall	0.8691
max_ap	0.952

Binary Classification

Dataset: pair-class-test
Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.8703
cosine_accuracy_threshold	0.8251
cosine_f1	0.8935
cosine_f1_threshold	0.8084
cosine_precision	0.8866
cosine_recall	0.9005
cosine_ap	0.9547
dot_accuracy	0.8703
dot_accuracy_threshold	0.8251
dot_f1	0.8935
dot_f1_threshold	0.8084
dot_precision	0.8866
dot_recall	0.9005
dot_ap	0.9547
manhattan_accuracy	0.8703
manhattan_accuracy_threshold	9.1812
manhattan_f1	0.8912
manhattan_f1_threshold	9.1812
manhattan_precision	0.9032
manhattan_recall	0.8796
manhattan_ap	0.9546
euclidean_accuracy	0.8703
euclidean_accuracy_threshold	0.5914
euclidean_f1	0.8935
euclidean_f1_threshold	0.619
euclidean_precision	0.8866
euclidean_recall	0.9005
euclidean_ap	0.9547
max_accuracy	0.8703
max_accuracy_threshold	9.1812
max_f1	0.8935
max_f1_threshold	9.1812
max_precision	0.9032
max_recall	0.9005
max_ap	0.9547

Training Details

Training Dataset

Unnamed Dataset

Size: 2,836 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: 15.88 tokens
max: 66 tokens

0: ~45.70%
1: ~54.30%

min: 5 tokens
mean: 15.82 tokens
max: 63 tokens

	sentence1	label	sentence2
type	string	int	string
details	min: 6 tokens mean: 15.88 tokens max: 66 tokens	0: ~45.70% 1: ~54.30%	min: 5 tokens mean: 15.82 tokens max: 63 tokens

Samples:

sentence1	label	sentence2
`What are the symptoms of diabetes?`	`1`	`What are the indicators of diabetes?`
`What is the speed of light?`	`1`	`At what speed does light travel?`
`Eager inventory processing loads the entire inventory list immediately and returns it, while lazy inventory processing applies the processing steps on-the-fly when browsing through the list.`	`1`	`Inventory processing that is done eagerly loads the entire inventory right away and provides the result, whereas lazy inventory processing performs the operations as it goes through the list.`

Loss: OnlineContrastiveLoss

Evaluation Dataset

Unnamed Dataset

Size: 316 evaluation samples
Columns: sentence1, label, and sentence2
Approximate statistics based on the first 316 samples:
sentence1 label sentence2
type string int string
details
min: 6 tokens
mean: 16.37 tokens
max: 98 tokens

0: ~39.56%
1: ~60.44%

min: 4 tokens
mean: 15.89 tokens
max: 98 tokens

	sentence1	label	sentence2
type	string	int	string
details	min: 6 tokens mean: 16.37 tokens max: 98 tokens	0: ~39.56% 1: ~60.44%	min: 4 tokens mean: 15.89 tokens max: 98 tokens

Samples:

sentence1	label	sentence2
`How many planets are in the solar system?`	`1`	`Number of planets in the solar system`
`What are the symptoms of pneumonia?`	`0`	`What are the symptoms of bronchitis?`
`What is the boiling point of sulfur?`	`0`	`What is the melting point of sulfur?`

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: 6
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: 6
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	-	-	0.8066	-
0.2247	10	1.6271	-	-	-
0.4494	20	1.0316	-	-	-
0.6742	30	0.7502	-	-	-
0.8989	40	0.691	-	-	-
0.9888	44	-	0.7641	0.9368	-
1.1236	50	0.732	-	-	-
1.3483	60	0.532	-	-	-
1.5730	70	0.389	-	-	-
1.7978	80	0.2507	-	-	-
2.0	89	-	0.6496	0.9516	-
2.0225	90	0.4147	-	-	-
2.2472	100	0.2523	-	-	-
2.4719	110	0.1588	-	-	-
2.6966	120	0.1168	-	-	-
2.9213	130	0.1793	-	-	-
2.9888	133	-	0.6431	0.9547	-
3.1461	140	0.2062	-	-	-
3.3708	150	0.109	-	-	-
3.5955	160	0.0631	-	-	-
3.8202	170	0.0588	-	-	-
4.0	178	-	0.6676	0.9512	-
4.0449	180	0.1865	-	-	-
4.2697	190	0.0303	-	-	-
4.4944	200	0.0301	-	-	-
4.7191	210	0.0416	-	-	-
4.9438	220	0.028	-	-	-
4.9888	222	-	0.6770	0.9518	-
5.1685	230	0.0604	-	-	-
5.3933	240	0.0129	-	-	-
5.6180	250	0.0747	-	-	-
5.8427	260	0.0069	-	-	-
5.9326	264	-	0.6755	0.9520	0.9547

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",
}