Thanks for the great work! 馃檶 Could you please share the approximate GPU hours or wall-clock time for Pre-training (700K steps, batch 512) and Post-training (60K steps, batch 128). This would be very helpful for replication planning. Thanks in advance!
Thanks for the great work! 馃檶 Could you please share the approximate GPU hours or wall-clock time for Pre-training (700K steps, batch 512) and Post-training (60K steps, batch 128). This would be very helpful for replication planning. Thanks in advance!