Is there any way to configure Weights and Biases when using multi-node with axolotl. What are the best practices for logging distributed training .
Situation: axolotl only see the main communication node. axolotl config wandb part:
wandb_project: gemma-3-cpt-pretraining
wandb_entity:
wandb_watch:
wandb_name: gemma-3-4b-cpt-100k
wandb_log_model:
resource: [Log distributed training experiments](Log distributed training experiments | Weights & Biases Documentation "Log distributed training experiments
(Log distributed training experiments | Weights & Biases Documentation)")
