[Code scan] GDI loop DAG crashes on missing Machine and Resources imports

Source: Codex global repository scan of deepmodeling/dpti at commit b719828e7eeb571bd26411430197cd74ea78e38c.
Project: https://github.com/orgs/deepmodeling/projects/27

Problem
`dpti_gdi_loop_md()` calls `Machine.load_from_dict()` and `Resources.load_from_dict()`, but `dpti/dags/dp_ti_gdi.py` only imports `Submission` and `Task` from `dpdispatcher.submission`. The loop DAG will fail with `NameError` when it reaches MD submission.

Code references
https://github.com/deepmodeling/dpti/blob/b719828e7eeb571bd26411430197cd74ea78e38c/dpti/dags/dp_ti_gdi.py#L19
https://github.com/deepmodeling/dpti/blob/b719828e7eeb571bd26411430197cd74ea78e38c/dpti/dags/dp_ti_gdi.py#L142
https://github.com/deepmodeling/dpti/blob/b719828e7eeb571bd26411430197cd74ea78e38c/dpti/dags/dp_ti_gdi.py#L143

Reproduction
Trigger the GDI loop DAG with valid `mdata`, `submission_dict`, and two task dictionaries. The MD task should deserialize dpdispatcher objects, but currently `Machine` and `Resources` are undefined.

Expected result
The DAG module should import `Machine` and `Resources` or use a dpdispatcher API that does not require these missing names.

Open dpti/dags/dp_ti_gdi.py and inspect the imports near line 19 and the MD submission path around lines 142–143. Trigger dpti_gdi_loop_md() with valid mdata, submission_dict, and two task dictionaries; done means the MD task deserializes the dpdispatcher objects without a NameError.

Written by the indexing model from the issue text.