Llama-LoRa Project

Quality evaluation and results

  • Fine-tuned on OpenOrca Open Llama 3B
TaskVersionMetricValueStderr
hellaswag0acc0.4899±0.0050
acc_norm0.6506±0.0048
  • Fine-tuned on my custom dataset in russian language Open Llama 3B
TaskVersionMetricValueStderr
hellaswag0acc0.4818±0.0050
acc_norm0.6377±0.0048
  • Fine-tuned on mixed dataset (one part – Orca, second part – custom dataset)
TaskVersionMetricValueStderr
hellaswag0acc0.4618±0.0050
acc_norm0.6019±0.0049
  • Fine-tuned on “wmt19”, ‘ru-en’ dataset
TaskVersionMetricValueStderr
hellaswag0acc0.4817±0.0050
acc_norm0.6362±0.0048
Andrii Semenov
Andrii Semenov
ML Research Scientist

My research interests include Natural Language Processing, Computer Vision and Stochastic Optimization.