Thanks very much for your work.
May I ask how long is the training time? How long is it for training the reward model on 156k dataset, then low long is the training of the 7B policy model on 130k dataset?
And it seems you don't cold start training the 7B model, directly applies RL on it? Will cold start help better?
Thanks very much for your work.
May I ask how long is the training time? How long is it for training the reward model on 156k dataset, then low long is the training of the 7B policy model on 130k dataset?
And it seems you don't cold start training the 7B model, directly applies RL on it? Will cold start help better?