metadata
license: apache-2.0
LLM360 Research Suite: K2 Loss Spike 1
We encountered two major loss spikes while training K2.
- The first loss spike occurred after 160 checkpoints and lasted over ~34 checkpoints. We restarted training at checkpoint 160 and training returned to normal.
- The second loss spike occurred after restarting training to fix the first loss spike at checkpoint 186 and lasted from ~8 checkpoints.
- For every spike checkpoint, we also uploaded the corresponding normal checkpoint for easy comparison. You could find different checkpoints in different branches.
We are releasing these checkpoints so others can study this interesting phenomena in large model training.
Purpose
Loss spikes are still a relatively unknown phenomena. By making these spikes and associated training details available, we hope others use these artifacts to further the worlds knowledge on this topic.
First 10 Checkpoints
| Checkpoints | |
|---|---|
| Checkpoint 160 | Checkpoint 170 |
| Checkpoint 162 | Checkpoint 172 |
| Checkpoint 164 | Checkpoint 174 |
| Checkpoint 166 | Checkpoint 176 |
| Checkpoint 168 | Checkpoint 178 |
[to find all branches: git branch -a]
Loss Spike's on the LLM360 Evaluation Suite
View all the evaluations on our Weights & Biases here
About the LLM360 Research Suite
The LLM360 Research Suite is a comprehensive set of large language model (LLM) artifacts from Amber, CrystalCoder, and K2 for academic and industry researchers to explore LLM training dynamics. Additional resources can be found at llm360.ai.
Citation
BibTeX:
@misc{
title={LLM360-K2-65B: Scaling Up Open and Transparent Language Models},
author={The LLM360 Team},
year={2024},
}