Add wf name to config; Better error format (#166)

Signed-off-by: zjgemi <liuxin_zijian@163.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Han Wang <92130845+wanghan-iapcm@users.noreply.github.com>

dpgen2/entrypoint/args.py

@@ -437,6 +437,7 @@ def submit_args(default_step_config=normalize_step_dict({})):
     doc_train = "The configuration for training"
     doc_explore = "The configuration for exploration"
     doc_fp = "The configuration for FP"
+    doc_name = "The workflow name, 'dpgen' for default"
 
     return (
         dflow_conf_args()
@@ -479,6 +480,7 @@ def submit_args(default_step_config=normalize_step_dict({})):
                 doc=doc_explore,
             ),
             Argument("fp", dict, [], [variant_fp()], optional=False, doc=doc_fp),
+            Argument("name", str, optional=True, default="dpgen", doc=doc_name),
         ]
     )
 

dpgen2/entrypoint/submit.py

@@ -718,7 +718,7 @@ def submit_concurrent_learning(
         # wf_config["inputs"]["do_finetune"] = False
         # finetune will not be done again if the old process is reused.
 
-    wf = Workflow(name="dpgen")
+    wf = Workflow(name=wf_config["name"])
 
     if wf_config["inputs"].get("do_finetune", False):
         assert finetune_step is not None

dpgen2/exploration/task/customized_lmp_template_task_group.py

@@ -180,14 +180,21 @@ def _make_customized_task_group(
                     # run shell command with dpgen2.utils.run_command
                     ret, out, err = run_command(ss, shell=True)
                     if ret != 0:
+                        logging.error(
+                            "".join(
+                                (
+                                    f"customized shell command {ss} failed with return code {ret}\n",
+                                    "out msg",
+                                    out,
+                                    "\n",
+                                    "err msg: ",
+                                    err,
+                                    "\n",
+                                )
+                            )
+                        )
                         raise FatalError(
-                            f"customized shell command {ss} failed with return code {ret}\n",
-                            "out msg",
-                            out,
-                            "\n",
-                            "err msg",
-                            err,
-                            "\n",
+                            f"customized shell command {ss} failed with return code {ret}"
                         )
                 # loop over all pattern matched result dirs
                 for ff in [

dpgen2/fp/gaussian.py

@@ -1,4 +1,5 @@
 """Prep and Run Gaussian tasks."""
+import logging
 from typing import (
     Any,
     List,
@@ -155,9 +156,20 @@ def run_task(
         command = " ".join([command, gaussian_input_name])
         ret, out, err = run_command(command, shell=True)
         if ret != 0:
-            raise TransientError(
-                "gaussian failed\n", "out msg", out, "\n", "err msg", err, "\n"
+            logging.error(
+                "".join(
+                    (
+                        "gaussian failed\n",
+                        "out msg: ",
+                        out,
+                        "\n",
+                        "err msg: ",
+                        err,
+                        "\n",
+                    )
+                )
             )
+            raise TransientError("gaussian failed")
         # convert the output to deepmd/npy format
         sys = dpdata.LabeledSystem(gaussian_output_name, fmt="gaussian/log")
         sys.to("deepmd/npy", out_name)

dpgen2/fp/vasp.py

@@ -1,3 +1,4 @@
+import logging
 from pathlib import (
     Path,
 )
@@ -132,9 +133,12 @@ def run_task(
         command = " ".join([command, ">", log_name])
         ret, out, err = run_command(command, shell=True)
         if ret != 0:
-            raise TransientError(
-                "vasp failed\n", "out msg", out, "\n", "err msg", err, "\n"
+            logging.error(
+                "".join(
+                    ("vasp failed\n", "out msg: ", out, "\n", "err msg: ", err, "\n")
+                )
             )
+            raise TransientError("vasp failed")
         # convert the output to deepmd/npy format
         sys = dpdata.LabeledSystem("OUTCAR")
         sys.to("deepmd/npy", out_name)

dpgen2/op/run_dp_train.py

@@ -1,5 +1,6 @@
 import glob
 import json
+import logging
 import os
 import shutil
 from pathlib import (
@@ -207,9 +208,20 @@ def clean_before_quit():
             ret, out, err = run_command(command)
             if ret != 0:
                 clean_before_quit()
-                raise FatalError(
-                    "dp train failed\n", "out msg", out, "\n", "err msg", err, "\n"
+                logging.error(
+                    "".join(
+                        (
+                            "dp train failed\n",
+                            "out msg: ",
+                            out,
+                            "\n",
+                            "err msg: ",
+                            err,
+                            "\n",
+                        )
+                    )
                 )
+                raise FatalError("dp train failed")
             fplog.write("#=================== train std out ===================\n")
             fplog.write(out)
             fplog.write("#=================== train std err ===================\n")
@@ -222,9 +234,20 @@ def clean_before_quit():
             ret, out, err = run_command(["dp", "freeze", "-o", "frozen_model.pb"])
             if ret != 0:
                 clean_before_quit()
-                raise FatalError(
-                    "dp freeze failed\n", "out msg", out, "\n", "err msg", err, "\n"
+                logging.error(
+                    "".join(
+                        (
+                            "dp freeze failed\n",
+                            "out msg: ",
+                            out,
+                            "\n",
+                            "err msg: ",
+                            err,
+                            "\n",
+                        )
+                    )
                 )
+                raise FatalError("dp freeze failed")
             fplog.write("#=================== freeze std out ===================\n")
             fplog.write(out)
             fplog.write("#=================== freeze std err ===================\n")

dpgen2/op/run_lmp.py

@@ -1,4 +1,5 @@
 import json
+import logging
 import os
 import random
 import re
@@ -151,17 +152,22 @@ def execute(
             command = " ".join([command, "-i", lmp_input_name, "-log", lmp_log_name])
             ret, out, err = run_command(command, shell=True)
             if ret != 0:
-                raise TransientError(
-                    "lmp failed\n",
-                    "command was",
-                    command,
-                    "out msg",
-                    out,
-                    "\n",
-                    "err msg",
-                    err,
-                    "\n",
+                logging.error(
+                    "".join(
+                        (
+                            "lmp failed\n",
+                            "command was: ",
+                            command,
+                            "out msg: ",
+                            out,
+                            "\n",
+                            "err msg: ",
+                            err,
+                            "\n",
+                        )
+                    )
                 )
+                raise TransientError("lmp failed")
 
         ret_dict = {
             "log": work_dir / lmp_log_name,