Merge branch 'deepmodeling:master' into master

zhaiwenxi · web-flow · commit 659d0900d4f8 · 2026-06-26T09:15:31.000+08:00
diff --git a/.github/workflows/build_cc.yml b/.github/workflows/build_cc.yml
@@ -27,7 +27,7 @@ jobs:
           - variant: clang
             dp_variant: clang
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
       - uses: actions/setup-python@v6
         with:
           python-version: "3.11"
diff --git a/.github/workflows/build_wheel.yml b/.github/workflows/build_wheel.yml
@@ -51,7 +51,7 @@ jobs:
             platform_id: manylinux_aarch64
             dp_variant: cpu
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
         with:
           # https://github.com/pypa/setuptools_scm/issues/480
           fetch-depth: 0
@@ -76,7 +76,7 @@ jobs:
     name: Build source distribution
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
         with:
           fetch-depth: 0
       - name: Build sdist
@@ -123,7 +123,7 @@ jobs:
           large-packages: true
           swap-storage: true
           docker-images: true
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
       - uses: actions/download-artifact@v8
         with:
           path: source/install/docker/dist
diff --git a/.github/workflows/codeql.yml b/.github/workflows/codeql.yml
@@ -30,7 +30,7 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v6
+        uses: actions/checkout@v7
       - uses: actions/setup-python@v6
         with:
           python-version: "3.11"
diff --git a/.github/workflows/copilot-setup-steps.yml b/.github/workflows/copilot-setup-steps.yml
@@ -30,7 +30,7 @@ jobs:
     # If you do not check out your code, Copilot will do this for you.
     steps:
       - name: Checkout code
-        uses: actions/checkout@v6
+        uses: actions/checkout@v7
 
       - name: Set up Python
         uses: actions/setup-python@v6
diff --git a/.github/workflows/package_c.yml b/.github/workflows/package_c.yml
@@ -35,7 +35,7 @@ jobs:
           large-packages: true
           swap-storage: true
           docker-images: true
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
         with:
           fetch-depth: 0
       - name: Package C library
@@ -63,7 +63,7 @@ jobs:
     needs: [build_c]
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
       - name: Download artifact
         uses: actions/download-artifact@v8
         with:
diff --git a/.github/workflows/test_cc.yml b/.github/workflows/test_cc.yml
@@ -35,7 +35,7 @@ jobs:
             enable_pytorch: false
             enable_paddle: true
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
       - uses: actions/setup-python@v6
         with:
           python-version: "3.11"
diff --git a/.github/workflows/test_cuda.yml b/.github/workflows/test_cuda.yml
@@ -23,7 +23,7 @@ jobs:
     #   options: --gpus all
     if: github.repository_owner == 'deepmodeling' && (github.event_name == 'pull_request' && github.event.label && github.event.label.name == 'Test CUDA' || github.event_name == 'workflow_dispatch' || github.event_name == 'merge_group')
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
       - uses: actions/setup-python@v6
         with:
           python-version: "3.11"
diff --git a/.github/workflows/test_python.yml b/.github/workflows/test_python.yml
@@ -22,7 +22,7 @@ jobs:
         python: ["3.10", "3.13"]
 
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
       - uses: actions/setup-python@v6
         with:
           python-version: ${{ matrix.python }}
@@ -48,7 +48,7 @@ jobs:
           UV_HTTP_TIMEOUT: 120
       - run: dp --version
       - name: Get durations from cache
-        uses: actions/cache@v5
+        uses: actions/cache@v6
         with:
           path: .test_durations
           # the key must never match, even when restarting workflows, as that
@@ -94,7 +94,7 @@ jobs:
     needs: testpython
     steps:
       - name: Get durations from cache
-        uses: actions/cache@v5
+        uses: actions/cache@v6
         with:
           path: .test_durations
           # key won't match during the first run for the given commit, but
diff --git a/.github/workflows/todo.yml b/.github/workflows/todo.yml
@@ -8,7 +8,7 @@ jobs:
     if: github.repository_owner == 'deepmodeling'
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v6
+      - uses: actions/checkout@v7
       - name: Run tdg-github-action
         uses: ribtoks/tdg-github-action@master
         with:
diff --git a/deepmd/pd/train/training.py b/deepmd/pd/train/training.py
@@ -81,6 +81,9 @@
 from deepmd.utils.data import (
     DataRequirementItem,
 )
+from deepmd.utils.finetune import (
+    warn_configuration_mismatch_during_finetune,
+)
 from deepmd.utils.path import (
     DPH5Path,
 )
@@ -520,9 +523,8 @@ def get_lr(lr_params: dict[str, Any]) -> BaseLR:
                     new_state_dict = {}
                     target_state_dict = self.wrapper.state_dict()
                     # pretrained_model
-                    pretrained_model = get_model_for_wrapper(
-                        state_dict["_extra_state"]["model_params"]
-                    )
+                    pretrained_model_params = state_dict["_extra_state"]["model_params"]
+                    pretrained_model = get_model_for_wrapper(pretrained_model_params)
                     pretrained_model_wrapper = ModelWrapper(pretrained_model)
                     pretrained_model_wrapper.set_state_dict(state_dict)
                     # update type related params
@@ -557,6 +559,25 @@ def collect_single_finetune_params(
                     ) -> None:
                         _new_fitting = _finetune_rule_single.get_random_fitting()
                         _model_key_from = _finetune_rule_single.get_model_branch()
+                        _input_model_params = (
+                            model_params["model_dict"][_model_key]
+                            if self.multi_task
+                            else model_params
+                        )
+                        _pretrained_model_params = (
+                            pretrained_model_params["model_dict"][_model_key_from]
+                            if "model_dict" in pretrained_model_params
+                            else pretrained_model_params
+                        )
+                        if (
+                            "descriptor" in _input_model_params
+                            and "descriptor" in _pretrained_model_params
+                        ):
+                            warn_configuration_mismatch_during_finetune(
+                                _input_model_params["descriptor"],
+                                _pretrained_model_params["descriptor"],
+                                _model_key_from,
+                            )
                         target_keys = [
                             i
                             for i in _random_state_dict.keys()
diff --git a/deepmd/pd/utils/finetune.py b/deepmd/pd/utils/finetune.py
@@ -8,6 +8,7 @@
 
 from deepmd.utils.finetune import (
     FinetuneRuleItem,
+    warn_descriptor_config_differences,
 )
 
 log = logging.getLogger(__name__)
@@ -57,6 +58,12 @@ def get_finetune_rule_single(
         random_fitting=new_fitting,
     )
     if change_model_params:
+        if "descriptor" in single_config and "descriptor" in single_config_chosen:
+            warn_descriptor_config_differences(
+                single_config["descriptor"],
+                single_config_chosen["descriptor"],
+                model_branch_chosen,
+            )
         trainable_param = {
             "descriptor": single_config.get("descriptor", {}).get("trainable", True),
             "fitting_net": single_config.get("fitting_net", {}).get("trainable", True),
diff --git a/deepmd/pt/train/training.py b/deepmd/pt/train/training.py
@@ -115,6 +115,9 @@
 from deepmd.utils.data import (
     DataRequirementItem,
 )
+from deepmd.utils.finetune import (
+    warn_configuration_mismatch_during_finetune,
+)
 
 if torch.__version__.startswith("2"):
     import torch._dynamo
@@ -793,9 +796,8 @@ def get_lr(lr_params: dict[str, Any]) -> BaseLR:
                     new_state_dict = {}
                     target_state_dict = self.wrapper.state_dict()
                     # pretrained_model
-                    pretrained_model = get_model_for_wrapper(
-                        state_dict["_extra_state"]["model_params"]
-                    )
+                    pretrained_model_params = state_dict["_extra_state"]["model_params"]
+                    pretrained_model = get_model_for_wrapper(pretrained_model_params)
                     pretrained_model_wrapper = ModelWrapper(pretrained_model)
                     pretrained_model_wrapper.load_state_dict(state_dict)
                     # update type related params
@@ -830,6 +832,25 @@ def collect_single_finetune_params(
                     ) -> None:
                         _new_fitting = _finetune_rule_single.get_random_fitting()
                         _model_key_from = _finetune_rule_single.get_model_branch()
+                        _input_model_params = (
+                            model_params["model_dict"][_model_key]
+                            if self.multi_task
+                            else model_params
+                        )
+                        _pretrained_model_params = (
+                            pretrained_model_params["model_dict"][_model_key_from]
+                            if "model_dict" in pretrained_model_params
+                            else pretrained_model_params
+                        )
+                        if (
+                            "descriptor" in _input_model_params
+                            and "descriptor" in _pretrained_model_params
+                        ):
+                            warn_configuration_mismatch_during_finetune(
+                                _input_model_params["descriptor"],
+                                _pretrained_model_params["descriptor"],
+                                _model_key_from,
+                            )
                         target_keys = [
                             i
                             for i in _random_state_dict.keys()
diff --git a/deepmd/pt/utils/finetune.py b/deepmd/pt/utils/finetune.py
@@ -14,6 +14,7 @@
 )
 from deepmd.utils.finetune import (
     FinetuneRuleItem,
+    warn_descriptor_config_differences,
 )
 from deepmd.utils.model_branch_dict import (
     get_model_dict,
@@ -69,6 +70,12 @@ def get_finetune_rule_single(
         random_fitting=new_fitting,
     )
     if change_model_params:
+        if "descriptor" in single_config and "descriptor" in single_config_chosen:
+            warn_descriptor_config_differences(
+                single_config["descriptor"],
+                single_config_chosen["descriptor"],
+                model_branch_chosen,
+            )
         trainable_param = {
             "descriptor": single_config.get("descriptor", {}).get("trainable", True),
             "fitting_net": single_config.get("fitting_net", {}).get("trainable", True),
diff --git a/deepmd/pt_expt/train/training.py b/deepmd/pt_expt/train/training.py
@@ -70,6 +70,9 @@
 from deepmd.utils.data_system import (
     DeepmdDataSystem,
 )
+from deepmd.utils.finetune import (
+    warn_configuration_mismatch_during_finetune,
+)
 from deepmd.utils.path import (
     DPPath,
 )
@@ -1121,6 +1124,29 @@ def _make_sample(
                             model_with_new_type_stat=model_with_new_type_stat,
                         )
 
+                for model_key in self.model_keys:
+                    finetune_rule = finetune_links[model_key]
+                    _model_key_from = finetune_rule.get_model_branch()
+                    input_model_params = (
+                        model_params["model_dict"][model_key]
+                        if self.multi_task
+                        else model_params
+                    )
+                    branch_pretrained_model_params = (
+                        pretrained_model_params["model_dict"][_model_key_from]
+                        if "model_dict" in pretrained_model_params
+                        else pretrained_model_params
+                    )
+                    if (
+                        "descriptor" in input_model_params
+                        and "descriptor" in branch_pretrained_model_params
+                    ):
+                        warn_configuration_mismatch_during_finetune(
+                            input_model_params["descriptor"],
+                            branch_pretrained_model_params["descriptor"],
+                            _model_key_from,
+                        )
+
                 # Selective weight copy (per-branch key remapping)
                 pretrained_state = pretrained_wrapper.state_dict()
                 target_state = self._unwrapped.state_dict()
diff --git a/deepmd/utils/finetune.py b/deepmd/utils/finetune.py
diff --git a/source/tests/common/test_finetune_utils.py b/source/tests/common/test_finetune_utils.py
