About the QAT code

[JingdiZhou]

Hello , I noticed that your contribution has contained QAT in qat.py , but when I read the code of train.py ,I found there is no real quantization of each model. For example, when I using command

python qat.py --algo a2c --env BreakoutNoFrameskip-v4 -q 7 --quant-delay 5000000 -n 10000000

I found that in a2c.py w_bits and act_bits actually haven't been used in other places ,which means the training process is actually without quantization . Therefore, I want to know if I misunderstood or confirmed that there was no QAT？

    def __init__(self, policy, env, gamma=0.99, n_steps=5, vf_coef=0.25, ent_coef=0.01, max_grad_norm=0.5,
                 learning_rate=7e-4, alpha=0.99, epsilon=1e-5, lr_schedule='constant', verbose=0,
                 tensorboard_log=None, _init_setup_model=True, w_bits=None, act_bits=None, quant_train=None, quant_delay=None, policy_kwargs=None,
                 full_tensorboard_log=False, seed=None, n_cpu_tf_sess=None):