Chi Zhang

You need to normalize the data first.

The best choice is actually investing anything for only 1 stock with highest return if we know the future. The reason that we split our investment is to avoid risk...

Maybe you want to take a look at the input scale. The pretrained model assumes the input is close/open ratio and normalize with (x - 1) / 100.

Yes. This is just a course project and there are a lot of practical issues regarding the environment. I suggest you look at other environment like this one [https://github.com/wassname/rl-portfolio-management](https://github.com/wassname/rl-portfolio-management).

> > Shapes of q, k, v are torch.Size([2, 8, 2048, 64]) in [b, h, n, d]. > > Currently, the backward on V100 isn't well parallelised. You will get...