fine_tuned_resume_matcher

Resume Matcher Transformer

A fine-tuned sentence transformer model based on sentence-transformers/all-MiniLM-L6-v2 optimized for comparing resumes with job descriptions.

Model Overview

This model transforms resumes and job descriptions into 384-dimensional embeddings that can be compared for semantic similarity, helping to identify the best candidates for a position.

Key Specifications

• Base Model: sentence-transformers/all-MiniLM-L6-v2
• Output Dimensions: 384
• Sequence Length: 256 tokens maximum
• Similarity Function: Cosine Similarity
• Pooling Strategy: Mean pooling

Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_mean_tokens': True})
  (2): Normalize()
)

Usage

# Install the required library
pip install -U sentence-transformers

from sentence_transformers import SentenceTransformer
from sklearn.metrics.pairwise import cosine_similarity

# Load the model
model = SentenceTransformer("path/to/model")

# Example job description
job_description = "Looking for a Python backend developer with Django experience."

# Example resumes
resume1 = "Experienced Python developer with Flask and Django skills."
resume2 = "Teacher with 5 years in classroom management experience."

# Generate embeddings
job_embedding = model.encode(job_description)
resume1_embedding = model.encode(resume1)
resume2_embedding = model.encode(resume2)

# Calculate similarity
similarity1 = cosine_similarity([job_embedding], [resume1_embedding])[0][0]
similarity2 = cosine_similarity([job_embedding], [resume2_embedding])[0][0]

print(f"Similarity with Resume 1: {similarity1:.4f}")
print(f"Similarity with Resume 2: {similarity2:.4f}")

Training Details

Dataset Information

• Size: 4 training samples
• Format: Pairs of text samples with similarity labels (0.0 = no match, 1.0 = match)
• Loss Function: CosineSimilarityLoss with MSELoss

Sample Training Data

Training Hyperparameters

• Training epochs: 4
• Batch size: 2
• Learning rate: 5e-05
• Optimizer: AdamW

<details><summary>View all hyperparameters</summary>

• per_device_train_batch_size: 2
• per_device_eval_batch_size: 2
• num_train_epochs: 4
• learning_rate: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1
• lr_scheduler_type: linear
• warmup_steps: 0
• seed: 42

</details>

Framework Versions

• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.6.0+cu124
• Python: 3.11.12

Citation

@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",
}