Update README.md

be5bf28 verified 4 months ago

21.2 kB

tags:
  - Machine-learning
  - Data-Science
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - dense
  - generated_from_trainer
  - dataset_size:134200
  - loss:MultipleNegativesRankingLoss
widget:
  - source_sentence: >-
      Currently, there seems to be breaking transformer news nearly every week
      with no sign of slowing.
    sentences:
      - >-
        Another active research aspect is concerned with improving the
        understanding of transformer-based models to further advance them.
      - >-
        For convenience, we choose to use a vector of unit length (its norm is
        1) and obtain this by dividing w by its norm, w ∥w∥ .
      - Measure the error (log Z Plog Z Q ) in bits.
  - source_sentence: >-
      Using the first three years of data, develop an appropriate ARIMA model
      and a procedure for these data.
    sentences:
      - >-
        Recall that the agent makes a decision at times determined by external
        events (or by other parts of the robot's control system).
      - >-
        A related fact is that in approximation in value space with multistep
        lookahead, J μ is the result of a step of Newton's method that starts at
        the function obtained by applying multiple value iterations to J.
      - c. Make one-step-ahead forecasts of the last 12 months.
  - source_sentence: >-
      The x-u-J notation is standard in deterministic optimal control textbooks
      (e.g., the classical books [AtF66] and [BrH75], noted earlier, as well as
      the more recent books by Stengel [Ste94], Kirk [Kir04], and Liberzon
      [Lib11]).
    sentences:
      - >-
        Sometimes the alternative notation p(j | i, u) is used for the
        transition probabilities. In the artificial intelligence literature, the
        focus is primarily on finitestate MDPs, particularly discounted and
        stochastic shortest path infinite horizon problems.
      - >-
        Research on artificial neural networks is as old as the digital
        computer.
      - >-
        Then the summations u P (u)f (u) should be written x P (x)f (u(x)). This
        means that P (u) is a finite sum of delta functions. This restriction
        guarantees that the mean and variance of u do exist, which is not
        necessarily the case for general P (u).
  - source_sentence: But how small can the error probability be?
    sentences:
      - >-
        The results described above allow us to offer the following service: if
        he tells us the properties of his channel, the desired rate R and the
        desired error probability p B , we can, after working out the relevant
        functions C, E r (R), and E sp (R), advise him that there exists a
        solution to his problem using a particular blocklength N ; indeed that
        almost any randomly chosen code with that blocklength should do the job.
      - >-
        If the problem space is complex, a few base-classifiers may be cascaded
        increasing the complexity at each stage.
      - >-
        Based on the fact that all these major implementations use
        normalization, it is clearly important, but why not just used standard
        batch normalization? Unfortunately, batch normalization is too memory
        intensive at our resolution. We have to come up with something that
        allows us to work with a few examples-that fit into our GPU memory with
        the two network graphs-but still works well.
  - source_sentence: >-
      If the weights are independent and the prior is taken as Gaussian, N (0,
      1/2λ)


      the MAP estimate minimizes the augmented error function


      where E is the usual classification or regression error (negative log
      likelihood).
    sentences:
      - >-
        This approach of removing unnecessary parameters is known as ridge
        regression in statistics.
      - >-
        If reselecting the test data doesn't help, you have other generalization
        problems.
      - >-
        These architectures are used across supervised, unsupervised, and
        reinforcement learning.
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - pearson_cosine
  - spearman_cosine
model-index:
  - name: SentenceTransformer
    results:
      - task:
          type: semantic-similarity
          name: Semantic Similarity
        dataset:
          name: val
          type: val
        metrics:
          - type: pearson_cosine
            value: null
            name: Pearson Cosine
          - type: spearman_cosine
            value: null
            name: Spearman Cosine
license: apache-2.0
datasets:
  - DigitalAsocial/ds-tb-17-g-sns-aml
language:
  - en
base_model:
  - sentence-transformers/all-MiniLM-L6-v2

SentenceTransformer

This is a sentence-transformers model trained. 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
Maximum Sequence Length: 256 tokens
Output Dimensionality: 384 dimensions
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': 256, 'do_lower_case': False, 'architecture': '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("sentence_transformers_model_id")
# Run inference
sentences = [
    'If the weights are independent and the prior is taken as Gaussian, N (0, 1/2λ)\n\nthe MAP estimate minimizes the augmented error function\n\nwhere E is the usual classification or regression error (negative log likelihood).',
    'This approach of removing unnecessary parameters is known as ridge regression in statistics.',
    'These architectures are used across supervised, unsupervised, and reinforcement learning.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.4685, 0.1523],
#         [0.4685, 1.0000, 0.1570],
#         [0.1523, 0.1570, 1.0000]])

Evaluation

Metrics

Semantic Similarity

Dataset: val
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	nan
spearman_cosine	nan

Training Details

Training Dataset

Training Data

The model was fine-tuned using 17 reference books in Data Science and Machine Learning, including: All source books were preprocessed using GROBID, an open-source tool for extracting and structuring text from PDF documents.
The raw PDF files were converted into structured text, segmented into sentences, and cleaned before being used for training.
This ensured consistent formatting and reliable sentence boundaries across the dataset.

Aßenmacher, Matthias. Multimodal Deep Learning. Self-published, 2023.
Bertsekas, Dimitri P. A Course in Reinforcement Learning. Arizona State University.
Boykis, Vicki. What are Embeddings. Self-published, 2023.
Bruce, Peter, and Andrew Bruce. Practical Statistics for Data Scientists: 50 Essential Concepts. O’Reilly Media, 2017.
Daumé III, Hal. A Course in Machine Learning. Self-published.
Deisenroth, Marc Peter, A. Aldo Faisal, and Cheng Soon Ong. Mathematics for Machine Learning. Cambridge University Press, 2020.
Devlin, Hannah, Guo Kunin, Xiang Tian. Seeing Theory. Self-published.
Gutmann, Michael U. Pen & Paper: Exercises in Machine Learning. Self-published.
Jung, Alexander. Machine Learning: The Basics. Springer, 2022.
Langr, Jakub, and Vladimir Bok. Deep Learning with Generative Adversarial Networks. Manning Publications, 2019.
MacKay, David J.C. Information Theory, Inference, and Learning Algorithms. Cambridge University Press, 2003.
Montgomery, Douglas C., Cheryl L. Jennings, and Murat Kulahci. Introduction to Time Series Analysis and Forecasting. 2nd Edition, Wiley, 2015.
Nilsson, Nils J. Introduction to Machine Learning: An Early Draft of a Proposed Textbook. Stanford University, 1996.
Prince, Simon J.D. Understanding Deep Learning. Draft Edition, 2024.
Shashua, Amnon. Introduction to Machine Learning. The Hebrew University of Jerusalem, 2008.
Sutton, Richard S., and Andrew G. Barto. Reinforcement Learning: An Introduction. 2nd Edition, MIT Press, 2018.
Alpaydin, Ethem. Introduction to Machine Learning. 3rd Edition, MIT Press, 2014.

⚠️ Note: Due to copyright restrictions, the full text of these books is not included in this repository. Only the fine-tuned model weights are shared.

Unnamed Dataset

Size: 134,200 training samples
Columns: sentence_0 and sentence_1
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1
type string string
details
min: 8 tokens
mean: 39.42 tokens
max: 256 tokens

min: 7 tokens
mean: 41.22 tokens
max: 256 tokens

	sentence_0	sentence_1
type	string	string
details	min: 8 tokens mean: 39.42 tokens max: 256 tokens	min: 7 tokens mean: 41.22 tokens max: 256 tokens

Samples:

sentence_0	sentence_1
`This equation is somewhat similar to what you have seen before (as a high-level simplification of equation 5.1), with some important differences.`	`The critic is trying to estimate the earth mover's distance, and looks for the maximum difference between the real (first term) and the generated (second term) distribution under different (valid) parametrizations of the f w function.`
`[2, p.173] Sketch the mutual information for this channel as a function of the input distribution p. Pick a convenient two-dimensional representation of p. The optimization routine must therefore take account of the possibility that, as we go up hill on I(X; Y ), we may run into the inequality constraints p i ≥ 0.`	`I(X; Y ) is a convex function of the channel parameters.`
`• Derive the AdaBoost algorithm.`	`• Understand the relationship between boosting decision stumps and linear classification.`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "gather_across_devices": false
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 6
fp16: True
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
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: 5e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1
num_train_epochs: 6
max_steps: -1
lr_scheduler_type: linear
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
bf16: False
fp16: True
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: False
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}
parallelism_config: None
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
project: huggingface
trackio_space_id: trackio
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
hub_revision: None
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: False
torch_compile_backend: None
torch_compile_mode: None
include_tokens_per_second: False
include_num_input_tokens_seen: no
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: True
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin
router_mapping: {}
learning_rate_mapping: {}

Training Logs

📈 Loss Curve (per epoch)

Epoch	Loss	Grad Norm
0.06	1.5921	19.76
0.30	1.2517	13.44
0.60	1.0856	13.96
0.95	0.9268	13.22
1.25	0.7569	14.07
1.61	0.6757	11.21
1.97	0.6409	12.76
2.32	0.5111	20.53
2.74	0.5059	14.88
3.10	0.3880	9.56
3.46	0.3792	9.78
3.87	0.3750	inf
4.11	0.3345	11.03
4.47	0.3271	13.21
4.83	0.3064	11.37
5.07	0.2752	19.23
5.36	0.2740	15.45
5.72	0.2773	6.55
5.96	0.2710	19.79
6.00	0.5610	— (final train loss)

Framework Versions

Python: 3.11.7
Sentence Transformers: 5.1.1
Transformers: 4.57.3
PyTorch: 2.5.1+cu121
Accelerate: 1.12.0
Datasets: 4.4.1
Tokenizers: 0.22.1

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",
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

If you use this model, please cite:

@misc{aghakhani2025synergsticrag,
  author       = {Danial Aghakhani Zadeh},
  title        = {Fine-tuned all-MiniLM-L6-v2 for Data Science RAG},
  year         = {2025},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/DigitalAsocial/all-MiniLM-L6-v2-ds-rag-s}}
}