docs: define HDF5 trajectory storage

docs/input.md

@@ -131,6 +131,10 @@ The {dargs:argument}`"stages"<explore[lmp]/stages>` defines the exploration stag
 
 The {dargs:argument}`"n_sample"<task_group[lmp-md]/n_sample>` tells the number of confgiruations randomly sampled from the set picked by {dargs:argument}`"conf_idx"<task_group[lmp-md]/conf_idx>` from {dargs:argument}`"configurations"<explore[lmp]/configurations>` for each exploration task. All configurations has the equal possibility to be sampled. The default value of `"n_sample"` is `null`, in this case all picked configurations are sampled. In the example, we have 3 samples for stage 0 task group 0 and 2 thermodynamic states (NVT, T=50 and 100K), then the task group has 3x2=6 NVT DPMD tasks.
 
+Set `explore.config.use_hdf5` to `true` to store LAMMPS trajectories and model deviations as dflow HDF5 dataset artifacts. LAMMPS still writes the standard text `traj.dump` and `model_devi.out` files inside each task; DPGEN2 serializes their data at the artifact boundary and reads it back transparently during selection. This avoids depending on the optional LAMMPS H5MD package and preserves the frame/model-deviation alignment expected by DPGEN2.
+
+This option is not the native LAMMPS `dump h5md` format. A custom H5MD dump cannot replace `traj.dump` in the built-in selector because its schema differs from the `lammps/dump` text consumed by dpdata. Use `extra_output_files` if a native H5MD file is needed as an additional output for another analysis tool.
+
 
 ### FP
 

dpgen2/entrypoint/args.py

@@ -359,7 +359,10 @@ def run_diffcsp_args():
     doc_gen_tasks = "Number of DiffCSP generation tasks"
     doc_gen_command = "Command for DiffCSP generation"
     doc_relax_group_size = "Group size for relaxation"
-    doc_use_hdf5 = "Use HDF5 to store trajs and model_devis"
+    doc_use_hdf5 = (
+        "Store trajectories and model deviations as dflow HDF5 dataset "
+        "artifacts; this is independent of LAMMPS dump h5md."
+    )
     return [
         Argument(
             "gen_tasks",

dpgen2/op/run_lmp.py

@@ -230,7 +230,11 @@ def lmp_args():
         doc_shuffle_models = "Randomly pick a model from the group of models to drive theexploration MD simulation"
         doc_head = "Select a head from multitask"
         doc_use_ele_temp = "Whether to use electronic temperature, 0 for no, 1 for frame temperature, and 2 for atomic temperature"
-        doc_use_hdf5 = "Use HDF5 to store trajs and model_devis"
+        doc_use_hdf5 = (
+            "Store the standard LAMMPS dump trajectory and model-deviation "
+            "values in dflow HDF5 dataset artifacts. This does not enable "
+            "LAMMPS's native dump h5md command."
+        )
         doc_extra_output_files = "Extra output file names, support wildcards"
         return [
             Argument("command", str, optional=True, default="lmp", doc=doc_lmp_cmd),

tests/op/test_run_lmp.py

@@ -67,6 +67,11 @@ def tearDown(self):
         if Path(self.task_name).is_dir():
             shutil.rmtree(self.task_name)
 
+    def test_hdf5_storage_option_is_normalized(self):
+        config = RunLmp.normalize_config({"use_hdf5": True})
+
+        self.assertTrue(config["use_hdf5"])
+
     @patch("dpgen2.op.run_lmp.run_command")
     def test_success(self, mocked_run):
         mocked_run.side_effect = [(0, "foo\n", "")]