Skip to content

cmtrhn/llm_recipes

Repository files navigation

Large Language Model Recipes

Here is a little showcase of few recipes on how to use Large Language Models (LLM) PyTorch Lightning is utilized for training module

install the requirements.txt

To train a translation model using Multi30K dataset and Seq2Seq Transformer

python -m spacy download de_core_news_sm
python -m spacy download en_core_web_sm
python train_a_translation_model.py --output 'address to your checkpoint output folder' --epochs 300

To get results from trained translation model

python train_a_translation_model.py --infer --checkpoint 'address to your checkpoint file'

To train a freshly defined BERT model with Masked Language Modelling and Next Sentence Prediction targets created from IMDB dataset

python pretrain_a_bert_model.py --output 'address to your checkpoint output folder' --epochs 300

To get results for NSP and MLM

python pretrain_a_bert_model.py --infer --checkpoint 'address to your checkpoint file'

About

Recipes for training LLM

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages