PyPI page
Home page
Author:
Martin Balázs Bánóczy
License:
MIT
Summary:
The library is built as a subclass of Huggingface’s Trainer, tailored for encoder-style fine-tuning of generative transformer models on sequence classification tasks. It enhances the standard training loop with features like a dedicated classification head, RAG (Retrieval-Augmented Generation) integrated predictions, and label-balanced batch sampling. The framework also includes advanced callbacks for improved experiment monitoring and real-time notifications.
Latest version:
1.0.6
Required dependencies:
datasets
|
faiss-cpu
|
gpustat
|
huggingface_hub
|
matplotlib
|
numpy
|
nvidia-htop
|
nvitop
|
openpyxl
|
pandas
|
regex
|
scikit-learn
|
seaborn
|
sentence-transformers
|
setuptools
|
slack_sdk
|
tabulate
|
tf_keras
|
tqdm
|
transformers
|
trl
|
unsloth
|
wandb
|
wheel
Downloads last day:
0
Downloads last week:
49
Downloads last month:
221