def train_with_losses_and_opt()

in phasic_policy_gradient/ppo.py [0:0]

• 18 lines of code
• 4 McCabe index (conditional complexity)

    def train_with_losses_and_opt(loss_keys, opt, **arrays):
        losses, diags = compute_losses(
            model,
            entcoef=entcoef,
            kl_penalty=kl_penalty,
            clip_param=clip_param,
            vfcoef=vfcoef,
            **arrays,
        )
        loss = sum([losses[k] * get_weight(k) for k in loss_keys])
        opt.zero_grad()
        loss.backward()
        tu.warn_no_gradient(model, "PPO")
        tu.sync_grads(params, grad_weight=grad_weight)
        diags = {k: v.detach() for (k, v) in diags.items()}
        opt.step()
        diags.update({f"loss_{k}": v.detach() for (k, v) in losses.items()})
        return diags