Datasets:

KrzTyb
/

fim-dataset

language:
  - code
tags:
  - code
  - c++
  - fill-in-the-middle
  - fim
  - autocomplete
  - code-completion
size_categories:
  - 10K<n<100K
task_categories:
  - text-generation
pretty_name: fim-dataset

fim-dataset

A Fill-in-the-Middle (FIM) dataset for code autocompletion.

Dataset Description

This dataset is designed for training code autocompletion models using the Fill-in-the-Middle (FIM) approach. The dataset contains code snippets formatted with FIM special tokens:

<fim_prefix>: Code before the completion point
<fim_suffix>: Code after the completion point
<fim_middle>: The code to be completed

Dataset Structure

Data Fields

text: The formatted FIM instruction containing prefix, suffix, and middle tokens

Data Splits

Split	Examples
train	42,922
validation	2,259

Usage

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("KrzTyb/fim-dataset")

# Access train and validation splits
train_data = dataset["train"]
val_data = dataset["validation"]

# Example: Print first training example
print(train_data[0]["text"])

Training

This dataset can be used to fine-tune code language models for autocompletion:

from transformers import AutoModelForCausalLM, AutoTokenizer, Trainer
from datasets import load_dataset

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-Coder-0.5B")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-Coder-0.5B")

# Add FIM tokens
fim_tokens = ["<fim_prefix>", "<fim_suffix>", "<fim_middle>", "<fim_pad>"]
tokenizer.add_special_tokens({"additional_special_tokens": fim_tokens})
model.resize_token_embeddings(len(tokenizer))

# Load dataset
dataset = load_dataset("KrzTyb/fim-dataset")

# Train (see full training script for details)
trainer = Trainer(
    model=model,
    train_dataset=dataset["train"],
    eval_dataset=dataset["validation"],
    # ... other training arguments
)
trainer.train()

Citation

If you use this dataset, please cite:

@dataset{KrzTyb_fim-dataset},
  title={fim-dataset},
  author={Dataset Creator},
  year={2025},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/datasets/KrzTyb/fim-dataset}}
}

License

Please specify the license for your dataset.