量化模型的时候出现不预期的问题

## 环境
1.系统环境：ubuntu1804
2.MegEngine版本：1.4.0
3.python版本：3.6.8

## 复现步骤
1. git clone https://github.com/MegEngine/Models.git
2. python3 train.py -a resnet18 -d /path/to/imagenet --mode normal 出现此问题

## 请提供关键的代码片段便于追查问题
1、train.py 和 finetune.py 同样出现此问题

## 请提供完整的日志及报错信息
Traceback (most recent call last):
File "finetune.py", line 315, in
main()
File "finetune.py", line 69, in main
train_proc(world_size, args)
File "finetune.py", line 159, in worker
data.RandomSampler(train_dataset, batch_size=cfg.BATCH_SIZE, drop_last=True)
File "/usr/local/python/lib/python3.6/site-packages/megengine/data/sampler.py", line 322, in __init__
self.sampler_iter = iter(self.sampler)
File "/usr/local/python/lib/python3.6/site-packages/megengine/data/sampler.py", line 100, in __iter__
return self.batch()
File "/usr/local/python/lib/python3.6/site-packages/megengine/data/sampler.py", line 146, in batch
if self.drop_last and len(batch_index[-1]) < self.batch_size:
IndexError: list index out of range

Reproduce the command from the issue in Models/train.py or finetune.py, then inspect the caller at finetune.py line 159 and MegEngine's data/sampler.py around lines 100 and 146. Confirm why RandomSampler reaches the empty batch case; done means the reported command no longer raises IndexError and the behavior is covered by an appropriate regression check.

Written by the indexing model from the issue text.