Merge branch 'deepmd-v3' into fix-dp-optim-parallel

wangzyphysics · Mar 29, 2024 · e6c9022 · e6c9022
2 parents 0e04dbe + 1c12011
commit e6c9022
Show file tree

Hide file tree

Showing 4 changed files with 866 additions and 4 deletions.
diff --git a/dpgen2/entrypoint/submit.py b/dpgen2/entrypoint/submit.py
@@ -475,10 +475,11 @@ def workflow_concurrent_learning(
     valid_data = config["inputs"]["valid_data_sys"]
     if valid_data is not None:
         valid_data_prefix = config["inputs"]["valid_data_prefix"]
+        valid_data = [valid_data] if isinstance(valid_data, str) else valid_data
+        assert isinstance(valid_data, list)
         if valid_data_prefix is not None:
             valid_data = [os.path.join(valid_data_prefix, ii) for ii in valid_data]
-        if isinstance(valid_data, str):
-            valid_data = expand_sys_str(valid_data)
+        valid_data = [expand_sys_str(ii) for ii in valid_data]
         valid_data = upload_artifact(valid_data)
     concurrent_learning_op = make_concurrent_learning_op(
         train_style,
@@ -543,10 +544,11 @@ def workflow_concurrent_learning(
         multi_init_data_idx = {}
         for k, v in multi_init_data.items():
             sys = v["sys"]
-            if isinstance(sys, str):
-                sys = expand_sys_str(sys)
+            sys = [sys] if isinstance(sys, str) else sys
+            assert isinstance(sys, list)
             if v["prefix"] is not None:
                 sys = [os.path.join(v["prefix"], ii) for ii in sys]
+            sys = [expand_sys_str(ii) for ii in sys]
             istart = len(init_data)
             init_data += sys
             iend = len(init_data)

diff --git a/examples/water/input_distill.json b/examples/water/input_distill.json
@@ -0,0 +1,198 @@
+{
+    "name": "water-distill",
+    "bohrium_config": {
+        "username": "",
+        "password": "",
+        "project_id": 1,
+        "_comment": "all"
+    },
+    "default_step_config": {
+        "template_config": {
+            "image": "",
+            "_comment": "all"
+        },
+        "_comment": "all"
+    },
+    "step_configs": {
+        "run_train_config": {
+            "template_config": {
+                "image": "",
+                "_comment": "all"
+            },
+            "executor": {
+                "type": "dispatcher",
+                "retry_on_submission_error": 10,
+                "image_pull_policy": "IfNotPresent",
+                "machine_dict": {
+                    "batch_type": "Bohrium",
+                    "context_type": "Bohrium",
+                    "remote_profile": {
+                        "input_data": {
+                            "job_type": "container",
+                            "platform": "ali",
+                            "scass_type": "1 * NVIDIA V100_16g"
+                        }
+                    }
+                }
+            },
+            "_comment": "all"
+        },
+        "run_explore_config": {
+            "template_config": {
+                "image": "",
+                "_comment": "all"
+            },
+            "continue_on_success_ratio": 0.80,
+            "executor": {
+                "type": "dispatcher",
+                "retry_on_submission_error": 10,
+                "image_pull_policy": "IfNotPresent",
+                "machine_dict": {
+                    "batch_type": "Bohrium",
+                    "context_type": "Bohrium",
+                    "remote_profile": {
+                        "input_data": {
+                            "job_type": "container",
+                            "platform": "ali",
+                            "scass_type": "1 * NVIDIA V100_16g"
+                        }
+                    }
+                }
+            },
+            "template_slice_config": {
+                "group_size": 5,
+                "pool_size": 1
+            },
+            "_comment": "all"
+        },
+        "run_fp_config": {
+            "template_config": {
+                "image": "",
+                "_comment": "all"
+            },
+            "continue_on_success_ratio": 0.80,
+            "executor": {
+                "type": "dispatcher",
+                "retry_on_submission_error": 10,
+                "image_pull_policy": "IfNotPresent",
+                "machine_dict": {
+                    "batch_type": "Bohrium",
+                    "context_type": "Bohrium",
+                    "remote_profile": {
+                        "input_data": {
+                            "job_type": "container",
+                            "platform": "ali",
+                            "scass_type": "1 * NVIDIA V100_16g"
+                        }
+                    }
+                }
+            },
+            "template_slice_config": {
+                "group_size": 500,
+                "pool_size": 1
+            },
+            "_comment": "all"
+        },
+        "_comment": "all"
+    },
+    "upload_python_packages": [
+        "/path/to/dpgen2"
+    ],
+    "inputs": {
+        "type_map": [
+            "O",
+            "H"
+        ],
+        "mixed_type": true,
+        "mass_map": [
+            16.0,
+            4.0
+        ],
+        "init_data_prefix": null,
+        "init_data_sys": [
+            "train_predict/data_0",
+            "train_predict/data_1",
+            "train_predict/data_2"
+        ],
+        "valid_data_sys": [
+            "valid_predict/data_3"
+        ],
+        "_comment": "all"
+    },
+    "train": {
+        "type": "dp",
+        "numb_models": 4,
+        "config": {
+            "init_model_policy": "yes",
+            "init_model_old_ratio": 0.90,
+            "init_model_numb_steps": 500000,
+            "init_model_start_lr": 1e-4,
+            "init_model_start_pref_e": 0.25,
+            "init_model_start_pref_f": 100,
+            "_comment": "all"
+        },
+        "template_script": "train.json",
+        "_comment": "all"
+    },
+    "explore": {
+        "type": "lmp",
+        "config": {
+            "command": "lmp -var restart 0"
+        },
+        "convergence": {
+            "type": "adaptive-lower",
+            "conv_tolerance": 0.005,
+            "_numb_candi_f": 3000,
+            "rate_candi_f": 0.15,
+            "level_f_hi": 0.5,
+            "n_checked_steps": 8,
+            "_command": "all"
+        },
+        "max_numb_iter": 16,
+        "fatal_at_max": false,
+        "configuration_prefix": null,
+        "configurations": [
+            {
+                "type": "file",
+                "files": [
+                    "init"
+                ],
+                "fmt": "deepmd/npy/mixed"
+            }
+        ],
+        "stages": [
+            [
+                {
+                    "type": "lmp-template",
+                    "lmp": "template.lammps",
+                    "trj_freq": 100,
+                    "revisions": {
+                        "V_NSTEPS": [
+                            100
+                        ],
+                        "V_TEMP": [
+                            330
+                        ],
+                        "V_DUMPFREQ": [
+                            200
+                        ]
+                    },
+                    "sys_idx": [
+                        0
+                    ],
+                    "n_sample": 4
+                }
+            ]
+        ],
+        "_comment": "all"
+    },
+    "fp": {
+        "type": "deepmd",
+        "task_max": 4000,
+        "run_config" : {
+            "teacher_model_path": "teacher_model.pt"
+        },
+        "inputs_config": {},
+        "_comment": "all"
+    }
+}