a question about inference

[duangtg]

Thank you for your excellent work and detailed presentation, but I have some questions about the code.
Why the time embedding sampled during training is [0, 99, 199, 299, ..., 999]， But when using the ddpm_steps function, the sampling steps used are [0, 100, 200, ..., 900]? Will this inconsistency affect the results? Also and what is the difference between the ddpm_steps and generalized_steps functions?
Thank you very much