PPO slow start

[felipemello1]

1. I tried using this https://github.com/pytorch/torchtune/blob/main/recipes/configs/mistral/7B_full_ppo_low_memory.yaml and gave up after 10min no first step, even after turning compile=False

2. we shouldn't be compiling the entire model. Instead, we should use the utility that compiles per layer, like here:

   torchtune/recipes/lora_dpo_single_device.py

   Line 296 in e9fd56a

   training.compile_model(model)

3. DPO distributed doesn't compile the model. Should it?

cc: @SalmanMohammadi