feat(pt): support zbl finetune #4849
Conversation
Updated the logic for transferring state dict items to correctly handle keys related to ZBL models by replacing '.models.0.' with '.' and ensuring '.models.1.' items are retained. This improves compatibility when loading pretrained models with different model key structures.
Introduces a compute_out_stat parameter to compute_or_load_stat methods in BaseAtomicModel, DPAtomicModel, LinearEnergyAtomicModel, and PairTabAtomicModel. This allows conditional computation of output statistics, improving flexibility and control over the statistics computation process.
|
Note Other AI code review bot(s) detectedCodeRabbit has detected other AI code review bot(s) in this pull request and will avoid duplicating their findings in the review comments. This may lead to a less comprehensive review. 📝 WalkthroughWalkthroughAdded an optional boolean parameter compute_or_load_out_stat (default True) to compute_or_load_stat across atomic models to allow skipping output-stat computation. Linear model output-stat handling was refactored (per-submodel output-stat removed, sampler wrapped to inject exclusions, apply_out_stat adjusted). Trainer finetune key-mapping was hardened for ZBL cases; tests updated. Changes
Sequence Diagram(s)sequenceDiagram
participant Caller
participant LinearModel
participant SubModel
participant Sampler
Caller->>LinearModel: compute_or_load_stat(sampled_func, stat_file_path, compute_or_load_out_stat)
LinearModel->>SubModel: compute_or_load_stat(sampled_func, stat_file_path, compute_or_load_out_stat=False)
Sampler->>LinearModel: samples (wrapped to inject pair/atom exclusion types)
alt compute_or_load_out_stat == True
LinearModel->>LinearModel: compute_or_load_out_stat(wrapped_sampler, stat_file_path)
LinearModel->>LinearModel: apply_out_stat(per-atom-type biases)
else
Note over LinearModel: Skip computing/loading output stats
end
Estimated code review effort🎯 4 (Complex) | ⏱️ ~30–45 minutes Suggested reviewers
✨ Finishing Touches
🧪 Generate unit tests
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. CodeRabbit Commands (Invoked using PR/Issue comments)Type Other keywords and placeholders
CodeRabbit Configuration File (
|
for more information, see https://pre-commit.ci
Enhanced Trainer to support fine-tuning ZBL models from standard models by handling key mapping and random state initialization. Added corresponding tests to verify ZBL fine-tuning behavior and ensure correct state dict transfer in test_training.py.
There was a problem hiding this comment.
Pull Request Overview
This PR adds support for ZBL (Ziegler-Biersack-Littmark) model fine-tuning from standard models, allowing atomic models to transition between different model types during training.
- Enhanced parameter handling during fine-tuning to support ZBL model initialization from standard models
- Modified atomic model statistics computation to provide optional control over output statistics loading
- Updated linear atomic model bias application logic to work directly with the model output
Reviewed Changes
Copilot reviewed 7 out of 7 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| source/tests/pt/test_training.py | Added test configuration and logic for ZBL fine-tuning from standard models |
| source/tests/pt/model/test_linear_atomic_model_stat.py | Simplified bias application verification in linear atomic model tests |
| deepmd/pt/train/training.py | Enhanced fine-tuning parameter collection to handle ZBL model branches and key mapping |
| deepmd/pt/model/atomic_model/pairtab_atomic_model.py | Added optional compute_out_stat parameter to control statistics computation |
| deepmd/pt/model/atomic_model/linear_atomic_model.py | Refactored bias application and statistics handling for linear atomic models |
| deepmd/pt/model/atomic_model/dp_atomic_model.py | Added conditional control for output statistics computation |
| deepmd/pt/model/atomic_model/base_atomic_model.py | Added compute_out_stat parameter to base interface |
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## devel #4849 +/- ##
==========================================
- Coverage 84.30% 84.29% -0.02%
==========================================
Files 702 702
Lines 68620 68647 +27
Branches 3572 3573 +1
==========================================
+ Hits 57850 57864 +14
- Misses 9630 9641 +11
- Partials 1140 1142 +2 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
deepmd/pt/train/training.py (1)
513-532: Consider improving variable naming and error messages per reviewer feedbackThe current implementation has two suggestions from previous reviews that remain valid:
- The variable name
use_random_initializationis misleading since it controls whether to use random initialization OR pretrained weights- The error message could be more descriptive to help users troubleshoot
🧹 Nitpick comments (1)
source/tests/pt/test_training.py (1)
255-266: Consider extracting common ZBL test setup to reduce duplicationThe ZBL configuration setup is duplicated between test classes. Consider extracting this into a helper method or base class to improve maintainability.
+ def _setup_zbl_config(self): + """Helper method to setup ZBL configuration for testing.""" + input_json_zbl = str(Path(__file__).parent / "water/zbl.json") + with open(input_json_zbl) as f: + self.config_zbl = json.load(f) + data_file = [str(Path(__file__).parent / "water/data/data_0")] + self.config_zbl["training"]["training_data"]["systems"] = data_file + self.config_zbl["training"]["validation_data"]["systems"] = data_file + self.config_zbl["model"] = deepcopy(model_zbl) + self.config_zbl["training"]["numb_steps"] = 1 + self.config_zbl["training"]["save_freq"] = 1 + def setUp(self) -> None: input_json = str(Path(__file__).parent / "water/se_atten.json") with open(input_json) as f: self.config = json.load(f) data_file = [str(Path(__file__).parent / "water/data/data_0")] self.config["training"]["training_data"]["systems"] = data_file self.config["training"]["validation_data"]["systems"] = data_file self.config["model"] = deepcopy(model_dpa1) self.config["training"]["numb_steps"] = 1 self.config["training"]["save_freq"] = 1 self.test_zbl_from_standard = True - - input_json_zbl = str(Path(__file__).parent / "water/zbl.json") - with open(input_json_zbl) as f: - self.config_zbl = json.load(f) - data_file = [str(Path(__file__).parent / "water/data/data_0")] - self.config_zbl["training"]["training_data"]["systems"] = data_file - self.config_zbl["training"]["validation_data"]["systems"] = data_file - self.config_zbl["model"] = deepcopy(model_zbl) - self.config_zbl["training"]["numb_steps"] = 1 - self.config_zbl["training"]["save_freq"] = 1 + self._setup_zbl_config()
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
deepmd/pt/train/training.py(1 hunks)source/tests/pt/test_training.py(3 hunks)
🧰 Additional context used
🧬 Code Graph Analysis (1)
source/tests/pt/test_training.py (5)
deepmd/pt/utils/finetune.py (1)
get_finetune_rules(82-203)deepmd/pt/entrypoints/main.py (1)
get_trainer(96-201)deepmd/dpmodel/output_def.py (2)
wrapper(45-72)wrapper(87-105)deepmd/pd/train/wrapper.py (1)
state_dict(199-203)deepmd/pt/train/training.py (1)
run(680-1170)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (29)
- GitHub Check: Build C++ (clang, clang)
- GitHub Check: Build C++ (rocm, rocm)
- GitHub Check: Build C++ (cuda120, cuda)
- GitHub Check: Build C++ (cpu, cpu)
- GitHub Check: Build C++ (cuda, cuda)
- GitHub Check: Build wheels for cp311-macosx_arm64
- GitHub Check: Build C library (2.18, libdeepmd_c.tar.gz)
- GitHub Check: Build C library (2.14, >=2.5.0,<2.15, libdeepmd_c_cu11.tar.gz)
- GitHub Check: Test Python (2, 3.12)
- GitHub Check: Test Python (4, 3.12)
- GitHub Check: Test Python (6, 3.9)
- GitHub Check: Test Python (4, 3.9)
- GitHub Check: Test Python (5, 3.9)
- GitHub Check: Test Python (1, 3.12)
- GitHub Check: Test Python (6, 3.12)
- GitHub Check: Test Python (3, 3.9)
- GitHub Check: Test Python (2, 3.9)
- GitHub Check: Test Python (5, 3.12)
- GitHub Check: Test Python (3, 3.12)
- GitHub Check: Test Python (1, 3.9)
- GitHub Check: Build wheels for cp311-macosx_x86_64
- GitHub Check: Build wheels for cp310-manylinux_aarch64
- GitHub Check: Build wheels for cp311-win_amd64
- GitHub Check: Build wheels for cp311-manylinux_x86_64
- GitHub Check: Build wheels for cp311-manylinux_x86_64
- GitHub Check: Analyze (c-cpp)
- GitHub Check: Analyze (python)
- GitHub Check: Test C++ (false)
- GitHub Check: Test C++ (true)
🔇 Additional comments (2)
source/tests/pt/test_training.py (2)
33-34: LGTM! Well-structured test attribute for ZBL finetuningThe addition of
test_zbl_from_standardas a class attribute is a clean way to enable ZBL finetuning tests selectively.
100-127: Test logic correctly verifies ZBL finetuning key mappingThe test implementation properly validates the key remapping logic for ZBL models:
- Maps ".models.0." keys back to their original form for comparison
- Skips ".models.1." keys as they should be randomly initialized
- Excludes bias/std keys from comparison as expected
The test provides good coverage of the finetuning logic introduced in the training module.
Revised and clarified the docstrings for compute_or_load_stat in both BaseAtomicModel and PairTabAtomicModel to better describe the function parameters and behavior. Updated parameter names and descriptions for improved consistency and readability.
There was a problem hiding this comment.
Actionable comments posted: 1
🔭 Outside diff range comments (1)
deepmd/pt/model/atomic_model/dp_atomic_model.py (1)
284-289: Propagatecompute_or_load_out_statthrough PT & PD wrappersThe public
compute_or_load_statin bothmake_model.pyfiles still only accept(sampled_func, stat_file_path)and drop the newcompute_or_load_out_statflag. Update them to match your atomic‐model API:• deepmd/pt/model/model/make_model.py
• deepmd/pd/model/model/make_model.pySuggested diff (apply to both files at the
compute_or_load_statdefinition):- def compute_or_load_stat( - self, - sampled_func, - stat_file_path: Optional[DPPath] = None, - ): - """Compute or load the statistics.""" - return self.atomic_model.compute_or_load_stat(sampled_func, stat_file_path) + def compute_or_load_stat( + self, + sampled_func, + stat_file_path: Optional[DPPath] = None, + compute_or_load_out_stat: bool = True, + ): + """Compute or load the statistics. Pass through `compute_or_load_out_stat`.""" + return self.atomic_model.compute_or_load_stat( + sampled_func, stat_file_path, compute_or_load_out_stat + )This ensures external callers can disable output‐stat computation via the public API.
♻️ Duplicate comments (2)
deepmd/pt/model/atomic_model/base_atomic_model.py (1)
369-386: Docstring update looks good and addresses prior feedbackThe description now clearly differentiates input vs output statistics and documents the new flag. This resolves the prior request to update the docs.
deepmd/pt/model/atomic_model/linear_atomic_model.py (1)
2-2: Nit: import ordering (duplicate of prior nitpick)There was a previous comment about the position of the functools import relative to typing. If you’re following a specific internal import order guideline, please align with it; otherwise, current ordering is acceptable.
🧹 Nitpick comments (7)
deepmd/pt/model/atomic_model/base_atomic_model.py (1)
362-367: Return type should beNone(notNoReturn) to match subclass implementationsSubclasses (e.g., DPAtomicModel, LinearEnergyAtomicModel) return
None. Annotating the base asNoReturn(a “never-returns” type) is incompatible with overriding methods that do return normally and may cause type-checker issues.Apply this diff:
def compute_or_load_stat( self, merged: Union[Callable[[], list[dict]], list[dict]], stat_file_path: Optional[DPPath] = None, - compute_or_load_out_stat: bool = True, -) -> NoReturn: + compute_or_load_out_stat: bool = True, +) -> None:Note: If
NoReturnbecomes unused after this change, we can remove it from the typing imports in a follow-up.deepmd/pt/model/atomic_model/dp_atomic_model.py (3)
291-297: Docstring nit: refer tosampled_funcconsistentlyUse the same parameter name as the function signature to avoid confusion.
-When `sampled` is provided, all the statistics parameters will be calculated (or re-calculated for update), +When `sampled_func` is provided, all the statistics parameters will be calculated (or re-calculated for update), -When `sampled` is not provided, it will check the existence of `stat_file_path`(s) +When `sampled_func` is not provided, it will check the existence of `stat_file_path`(s)
302-304: Docstring nit:stat_file_pathtype is a path, not a dictThe doc currently says “The dictionary of paths…”. The type is
Optional[DPPath].- stat_file_path - The dictionary of paths to the statistics files. + stat_file_path + The path to the statistics file(s).
313-325: Cache the sampler with an explicit bound
wrapped_samplertakes no arguments, so the cache will contain a single entry in practice. Still, explicitly settingmaxsize=1makes the intent clear and prevents accidental unbounded growth if arguments are added later.- @functools.lru_cache + @functools.lru_cache(maxsize=1) def wrapped_sampler():deepmd/pt/model/atomic_model/linear_atomic_model.py (3)
321-326: Avoid duplication: delegate to base implementation of apply_out_statThis override now matches BaseAtomicModel.apply_out_stat verbatim. Prefer delegating to reduce maintenance overhead and divergence risk.
def apply_out_stat( self, ret: dict[str, torch.Tensor], atype: torch.Tensor, ): """Apply the stat to each atomic output. @@ - out_bias, out_std = self._fetch_out_stat(self.bias_keys) - for kk in self.bias_keys: - # nf x nloc x odims, out_bias: ntypes x odims - ret[kk] = ret[kk] + out_bias[kk][atype] - return ret + return super().apply_out_stat(ret, atype)
471-476: Docstring nit: refer tosampled_funcconsistentlyMirror the function signature to avoid confusion.
-When `sampled` is provided, all the statistics parameters will be calculated (or re-calculated for update), +When `sampled_func` is provided, all the statistics parameters will be calculated (or re-calculated for update), -When `sampled` is not provided, it will check the existence of `stat_file_path`(s) +When `sampled_func` is not provided, it will check the existence of `stat_file_path`(s)
489-494: Docstring nit:stat_file_pathtype is a path, not a dictAlign the description with the actual type:
Optional[DPPath].- stat_file_path - The dictionary of paths to the statistics files. + stat_file_path + The path to the statistics file(s).
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (4)
deepmd/pt/model/atomic_model/base_atomic_model.py(2 hunks)deepmd/pt/model/atomic_model/dp_atomic_model.py(3 hunks)deepmd/pt/model/atomic_model/linear_atomic_model.py(4 hunks)deepmd/pt/model/atomic_model/pairtab_atomic_model.py(2 hunks)
🚧 Files skipped from review as they are similar to previous changes (1)
- deepmd/pt/model/atomic_model/pairtab_atomic_model.py
🧰 Additional context used
🧬 Code Graph Analysis (1)
deepmd/pt/model/atomic_model/linear_atomic_model.py (5)
deepmd/pt/model/atomic_model/base_atomic_model.py (3)
_fetch_out_stat(576-592)compute_or_load_out_stat(389-414)compute_or_load_stat(362-387)deepmd/pt/model/atomic_model/dp_atomic_model.py (2)
compute_or_load_stat(284-331)wrapped_sampler(314-324)deepmd/pd/model/model/make_model.py (1)
compute_or_load_stat(568-574)deepmd/pt/model/model/make_model.py (1)
compute_or_load_stat(572-578)deepmd/pt/utils/exclude_mask.py (2)
get_exclude_types(35-36)get_exclude_types(101-102)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (26)
- GitHub Check: Build C library (2.14, >=2.5.0,<2.15, libdeepmd_c_cu11.tar.gz)
- GitHub Check: Build C library (2.18, libdeepmd_c.tar.gz)
- GitHub Check: Build C++ (cuda, cuda)
- GitHub Check: Build C++ (rocm, rocm)
- GitHub Check: Build C++ (cpu, cpu)
- GitHub Check: Build C++ (cuda120, cuda)
- GitHub Check: Build wheels for cp311-macosx_x86_64
- GitHub Check: Build wheels for cp311-manylinux_x86_64
- GitHub Check: Build wheels for cp310-manylinux_aarch64
- GitHub Check: Build wheels for cp311-win_amd64
- GitHub Check: Build wheels for cp311-manylinux_x86_64
- GitHub Check: Test Python (3, 3.12)
- GitHub Check: Test Python (5, 3.9)
- GitHub Check: Test Python (6, 3.9)
- GitHub Check: Test Python (4, 3.9)
- GitHub Check: Test Python (2, 3.12)
- GitHub Check: Test Python (6, 3.12)
- GitHub Check: Test Python (3, 3.9)
- GitHub Check: Test Python (4, 3.12)
- GitHub Check: Test Python (1, 3.12)
- GitHub Check: Test Python (2, 3.9)
- GitHub Check: Test Python (5, 3.12)
- GitHub Check: Test Python (1, 3.9)
- GitHub Check: Test C++ (false)
- GitHub Check: Test C++ (true)
- GitHub Check: Analyze (c-cpp)
🔇 Additional comments (2)
deepmd/pt/model/atomic_model/dp_atomic_model.py (1)
284-307: Conditional computation of output stats is correctThe new
compute_or_load_out_statflag is properly surfaced and documented here, and the computation is guarded accordingly. This aligns with the PR objective.deepmd/pt/model/atomic_model/linear_atomic_model.py (1)
495-499: Good: sub-models skip output-stat computationPassing
compute_or_load_out_stat=Falseto sub-models avoids double-biasing. This aligns with the new design.
There was a problem hiding this comment.
Actionable comments posted: 1
🔭 Outside diff range comments (1)
deepmd/pt/model/atomic_model/base_atomic_model.py (1)
362-367: Harmonizecompute_or_load_statSignatures and Return TypesAll overrides of
compute_or_load_statmust match the base signature and return annotation. Change the base method’s return type fromNoReturntoNoneand introduce the newcompute_or_load_out_stat: bool = Trueparameter in every subclass override. No call-site changes are needed since the new parameter has a default.Files requiring updates (add the third parameter and change
-> NoReturnto-> None):
- deepmd/pt/model/atomic_model/base_atomic_model.py (lines 362–367)
- deepmd/pt/model/atomic_model/pairtab_atomic_model.py (lines 225–227)
- deepmd/pt/model/atomic_model/linear_atomic_model.py (lines 471–473)
- deepmd/pt/model/atomic_model/dp_atomic_model.py (lines 284–286)
- deepmd/pt/model/model/model.py (lines 29–31)
- deepmd/pt/model/model/spin_model.py (lines 344–346)
- deepmd/pt/model/model/make_model.py (lines 572–574)
- deepmd/pd/model/atomic_model/base_atomic_model.py (lines 365–367)
- deepmd/pd/model/atomic_model/dp_atomic_model.py (lines 329–331)
- deepmd/pd/model/model/model.py (lines 23–25)
- deepmd/pd/model/model/make_model.py (lines 568–570)
Apply this snippet to each definition:
- def compute_or_load_stat( - self, - …, - stat_file_path: Optional[DPPath] = None, -) -> NoReturn: + def compute_or_load_stat( + self, + …, + stat_file_path: Optional[DPPath] = None, + compute_or_load_out_stat: bool = True, +) -> None:Optionally, mark the base method as
@abstractmethod(imported fromabc) for clarity.
♻️ Duplicate comments (1)
deepmd/pt/model/atomic_model/base_atomic_model.py (1)
369-385: Docstring mismatches: fix parameter names and path description
- Use “merged” consistently instead of “sampled”.
- stat_file_path is a single DPPath, not “dictionary of paths”.
Apply this doc-only diff:
-Compute or load the statistics parameters of the model, -such as mean and standard deviation of descriptors or the energy bias of the fitting net. -When `sampled` is provided, all the statistics parameters will be calculated (or re-calculated for update), -and saved in the `stat_file_path`(s). -When `sampled` is not provided, it will check the existence of `stat_file_path`(s) -and load the calculated statistics parameters. +Compute or load the statistics parameters of the model, +such as mean and standard deviation of descriptors or the energy bias of the fitting net. +When `merged` is provided, all the statistics parameters will be calculated (or re-calculated for update), +and saved to the given `stat_file_path`. +When `merged` is not provided, it will check the existence of `stat_file_path` +and load the calculated statistics parameters. @@ -merged - The lazy sampled function to get data frames from different data systems. +merged + The lazy sampled function to get data frames from different data systems. @@ -stat_file_path - The dictionary of paths to the statistics files. +stat_file_path + The path to the statistics file. @@ -compute_or_load_out_stat : bool +compute_or_load_out_stat : bool Whether to compute the output statistics. If False, it will only compute the input statistics (e.g. mean and standard deviation of descriptors).
🧹 Nitpick comments (1)
deepmd/pt/model/atomic_model/pairtab_atomic_model.py (1)
232-248: Docstring: align names and path description with the signature and types
- Replace “sampled” with “merged” to match the parameter.
- stat_file_path is a single DPPath, not a “dictionary of paths”.
Apply this doc-only diff:
-Compute or load the statistics parameters of the model, -such as mean and standard deviation of descriptors or the energy bias of the fitting net. -When `sampled` is provided, all the statistics parameters will be calculated (or re-calculated for update), -and saved in the `stat_file_path`(s). -When `sampled` is not provided, it will check the existence of `stat_file_path`(s) -and load the calculated statistics parameters. +Compute or load the statistics parameters of the model, +such as mean and standard deviation of descriptors or the energy bias of the fitting net. +When `merged` is provided, all the statistics parameters will be calculated (or re-calculated for update), +and saved to the given `stat_file_path`. +When `merged` is not provided, it will check the existence of `stat_file_path` +and load the calculated statistics parameters. @@ -sampled_func - The lazy sampled function to get data frames from different data systems. +merged + The lazy sampled function to get data frames from different data systems. @@ -stat_file_path - The dictionary of paths to the statistics files. +stat_file_path + The path to the statistics file. @@ -compute_or_load_out_stat : bool +compute_or_load_out_stat : bool Whether to compute the output statistics. If False, it will only compute the input statistics (e.g. mean and standard deviation of descriptors).
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
deepmd/pt/model/atomic_model/base_atomic_model.py(1 hunks)deepmd/pt/model/atomic_model/pairtab_atomic_model.py(1 hunks)
🧰 Additional context used
🧬 Code Graph Analysis (1)
deepmd/pt/model/atomic_model/pairtab_atomic_model.py (2)
deepmd/utils/path.py (1)
DPPath(27-157)deepmd/pt/model/atomic_model/base_atomic_model.py (1)
compute_or_load_out_stat(389-414)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (20)
- GitHub Check: Build wheels for cp311-manylinux_x86_64
- GitHub Check: Build wheels for cp311-manylinux_x86_64
- GitHub Check: Build wheels for cp311-macosx_x86_64
- GitHub Check: Analyze (c-cpp)
- GitHub Check: Test Python (6, 3.12)
- GitHub Check: Test Python (6, 3.9)
- GitHub Check: Test Python (2, 3.12)
- GitHub Check: Test Python (4, 3.12)
- GitHub Check: Test Python (5, 3.9)
- GitHub Check: Test Python (5, 3.12)
- GitHub Check: Test Python (4, 3.9)
- GitHub Check: Test Python (3, 3.9)
- GitHub Check: Test Python (2, 3.9)
- GitHub Check: Test Python (1, 3.12)
- GitHub Check: Test Python (1, 3.9)
- GitHub Check: Test Python (3, 3.12)
- GitHub Check: Build C library (2.18, libdeepmd_c.tar.gz)
- GitHub Check: Build C library (2.14, >=2.5.0,<2.15, libdeepmd_c_cu11.tar.gz)
- GitHub Check: Test C++ (false)
- GitHub Check: Test C++ (true)
🔇 Additional comments (1)
deepmd/pt/model/atomic_model/pairtab_atomic_model.py (1)
250-251: LGTM: guarded invocation of output-stat computationConditionally computing output stats respects the new flag and is appropriate for PairTab (no input stats to handle here).
3 participants
Summary by CodeRabbit
New Features
Bug Fixes
Refactor
Tests