different tensor dimensions from original paper in generator

Hello, I noticed that the generator and the sampler have different dimensions than in the original paper.
For example, in sampler, `h0 = tf.reshape(linear(z, self.gf_dim*8*s_h16*s_w16, 'g_h0_lin'), [-1, s_h16, s_w16, self.gf_dim * 8])`, I think self.gf_dim should be multiplied by 16 in order to reproduce the dimensions in the original paper. Is there any reason behind multiplying it only by 8 and thus having only half of that dimension?

Thank you very much!

P.S. also referencing #214

Start with the sampler expression for h0 in the issue and compare its tensor dimensions with the original paper and the corresponding generator dimensions. Review the discussion referenced in #214 and determine whether the factor of 8 is intentional. Done means resolving the discrepancy with a documented rationale or an aligned implementation.

Written by the indexing model from the issue text.