GPT-NeoX 20B Pretraining
Created on November 4|Last edited on April 10
Comment
Note that the evaluation metrics are not logged for the entire training and underestimate the true performance across the board. The final values are:
Lambada: 0.720
HellaSwag: 0.535
PiQA: 0.779
MathQA: [to do]
PubMedQA: [to do]
Winogrande: 0.661
Training Metrics
Evaluation Metrics
Add a comment