WarmupDecayLR.warmup_num_steps must not be 0 or 1

[stas00]

When using WarmupDecayLR, either

• the config checker must ensure that warmup_num_steps is not 0 or 1, because it's a 1 / log(warmup_num_steps)
• or the code needs to be smart enough to handle these 2 cases internally, .e.g:

warmup_num_steps = max(2, warmup_num_steps)

Thank you.

Fix: #772

Error log from the test:

    def __init__(self,
                 optimizer: Optimizer,
                 warmup_min_lr: float = 0.0,
                 warmup_max_lr: float = 0.001,
                 warmup_num_steps: int = 1000,
                 last_batch_iteration: int = -1):
    
        self.optimizer = get_torch_optimizer(optimizer)
    
        self.min_lrs = self._format_param(self.optimizer, warmup_min_lr, "min_lr")
        self.max_lrs = self._format_param(self.optimizer, warmup_max_lr, "max_lr")
        self.delta_lrs = [big - small for big, small in zip(self.max_lrs, self.min_lrs)]
        self.warmup_num_steps = warmup_num_steps
>       self.inverse_log_warm_up = 1.0 / math.log(warmup_num_steps)
E       ZeroDivisionError: float division by zero

DeepSpeed/deepspeed/runtime/lr_schedules.py:710: ZeroDivisionError