gte-modernbert-base trained on Magpie Coder data

This is a sentence-transformers model finetuned from Alibaba-NLP/gte-modernbert-base on the magpie-qwen2.5-coder-pro-300_k-query-positive-pair dataset. It maps sentences & paragraphs to a 768-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: Alibaba-NLP/gte-modernbert-base
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • magpie-qwen2.5-coder-pro-300_k-query-positive-pair
• Language: en
• License: apache-2.0

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': 8192, 'do_lower_case': False}) with Transformer model: ModernBertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

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("lemon-mint/gte-modernbert-base-code-3")
# Run inference
sentences = [
    'Design a simple Python script to backup a database using SQL commands and save it to a file. This script should use PostgreSQL as the database management system.',
    'import subprocess\nimport os\nfrom datetime import datetime\n\ndef backup_database(db_name, db_user, db_host, db_port, backup_dir):\n    # Ensure the backup directory exists\n    if not os.path.exists(backup_dir):\n        os.makedirs(backup_dir)\n\n    # Create a timestamp for the backup file\n    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")\n    backup_file = os.path.join(backup_dir, f"{db_name}_backup_{timestamp}.sql")\n\n    # Construct the pg_dump command\n    command = [\n        "pg_dump",\n        "-U", db_user,\n        "-h", db_host,\n        "-p", str(db_port),\n        "-F", "c",  # Custom format, which is compressed\n        "-b",       # Include large objects\n        "-v",       # Verbose mode\n        "-f", backup_file,\n        db_name\n    ]\n\n    try:\n        # Run the pg_dump command\n        subprocess.run(command, check=True)\n        print(f"Backup successful: {backup_file}")\n    except subprocess.CalledProcessError as e:\n        print(f"An error occurred while backing up the database: {e}")\n\nif __name__ == "__main__":\n    # Database connection details\n    db_name = "your_database_name"\n    db_user = "your_database_user"\n    db_host = "localhost"  # or your database host\n    db_port = 5432         # default PostgreSQL port\n    backup_dir = "/path/to/your/backup/directory"\n\n    # Call the backup function\n    backup_database(db_name, db_user, db_host, db_port, backup_dir)\n',
    'def fibonacci_sequence(n):\n    # Initialize the first two Fibonacci numbers\n    fib_sequence = [0, 1]\n    \n    # Generate the Fibonacci sequence up to n\n    while True:\n        next_fib = fib_sequence[-1] + fib_sequence[-2]\n        if next_fib > n:\n            break\n        fib_sequence.append(next_fib)\n    \n    # If n is less than 1, return the sequence up to n\n    if n < 1:\n        return fib_sequence[:n+1]\n    \n    return fib_sequence\n\n# Example usage\nn = int(input("Enter a number to generate the Fibonacci sequence up to: "))\nsequence = fibonacci_sequence(n)\nprint(f"Fibonacci sequence up to {n}: {sequence}")\n',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Training Details

Training Dataset

magpie-qwen2.5-coder-pro-300_k-query-positive-pair

• Dataset: magpie-qwen2.5-coder-pro-300_k-query-positive-pair at 1c3e977
• Size: 78,926 training samples
• Columns: query and positive
• Approximate statistics based on the first 1000 samples:

  	query	positive
  type	string	string
  details	min: 8 tokens mean: 35.38 tokens max: 184 tokens	min: 3 tokens mean: 235.42 tokens max: 1097 tokens

• Samples:

  query	positive
  Can you provide a Python script that reads a CSV file and performs basic data analysis on it? The CSV file contains sales data with columns like Date, Product, Quantity, and Price. The analysis should include:	import pandas as pd # Load the CSV file into a DataFrame def load_data(file_path): try: data = pd.read_csv(file_path) return data except FileNotFoundError: print(f"Error: The file {file_path} was not found.") return None except pd.errors.EmptyDataError: print("Error: The file is empty.") return None except pd.errors.ParserError: print("Error: There was a problem parsing the file.") return None # Perform basic data analysis def analyze_data(data): if data is None: return # Display the first few rows of the DataFrame print("First few rows of the dataset:") print(data.head()) # Get a summary of the DataFrame print("\nSummary of the dataset:") print(data.info()) # Get basic statistics of the DataFrame print("\nBasic statistics of the dataset:") print(data.describe()) # Calculate total sales data['Total Sales'] = data['Quantity'] * data['Price'] t...
  How can I retrieve the virtual machine instances using the Alibaba Cloud SDK for Java?	import com.aliyuncs.DefaultAcsClient; import com.aliyuncs.IAcsClient; import com.aliyuncs.ecs.model.v20140526.DescribeInstancesRequest; import com.aliyuncs.ecs.model.v20140526.DescribeInstancesResponse; import com.aliyuncs.exceptions.ClientException; import com.aliyuncs.profile.DefaultProfile; public class DescribeInstancesExample { public static void main(String[] args) { // Step 1: Configure your Alibaba Cloud credentials String accessKeyId = "yourAccessKeyId"; String accessKeySecret = "yourAccessKeySecret"; String regionId = "yourRegionId"; // e.g., "cn-hangzhou" // Step 2: Create a profile and initialize the client DefaultProfile profile = DefaultProfile.getProfile(regionId, accessKeyId, accessKeySecret); IAcsClient client = new DefaultAcsClient(profile); // Step 3: Create a request to describe instances DescribeInstancesRequest request = new DescribeInstancesRequest(); request.setPageSize(10); /...
  I have a list of product reviews where the titles and body text of each review are separated by a period. How can I split these reviews into their respective titles and bodies in Python?	# Sample list of reviews reviews = [ "Great product. I really enjoyed using it!", "Not what I expected. It broke after a week.", "Excellent quality. Very satisfied with the purchase." ] # Function to split reviews into titles and bodies def split_reviews(reviews): split_reviews = [] for review in reviews: # Split the review into title and body parts = review.split('.', 1) # Split only at the first period if len(parts) == 2: title = parts[0].strip() # Remove any leading/trailing whitespace body = parts[1].strip() # Remove any leading/trailing whitespace split_reviews.append((title, body)) else: # Handle cases where there is no period in the review split_reviews.append((parts[0].strip(), '')) return split_reviews # Split the reviews split_reviews = split_reviews(reviews) # Print the results for title, body in split_reviews: print(f"Title: {title}") print(f"Bo...

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 64
• learning_rate: 2e-05
• num_train_epochs: 2
• warmup_ratio: 0.1
• bf16: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 8
• 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: 2e-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: 2
• 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: True
• 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: 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}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch
• 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: None
• hub_always_push: False
• 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
• 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
• eval_on_start: False
• use_liger_kernel: False
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.4.0
• Transformers: 4.48.1
• PyTorch: 2.1.0+cu118
• Accelerate: 1.3.0
• Datasets: 3.2.0
• Tokenizers: 0.21.0

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}
}