how to set timestep shift when train flux-dev lora?

I am using flux_train_network.py to train the FLUX-dev model with the options

"--timestep_sampling", "sigmoid", "--discrete_flow_shift", "3.0"

Additionally, I want to train using options like the following example:

end_timestep_shift: 0.7,
discrete_flow_shift: 3,
start_timestep_shift: 3

I want to use start and end timestep shifts like these options for training.
Is this currently provided as an option? If not, what modifications are needed?

Start with flux_train_network.py and its handling of timestep_sampling and discrete_flow_shift. Determine whether start_timestep_shift and end_timestep_shift are supported, and define the implementation scope needed for the requested training configuration. Done means the requested shift settings are supported or their absence and required changes are clearly documented.

Written by the indexing model from the issue text.