AssertionError: 'coefficient' is not defined in the energy loss config {'fn': 'mae'}.

[yxwang1215]

Hi,

When I tried to follow the tutorial to run "python main.py --mode train --config-yml configs/is2re/10k/schnet/schnet.yml", this Error happened.

2025-02-25 10:53:14 (INFO): Loading model: schnet
2025-02-25 10:53:14 (INFO): Loaded SchNetWrap with 541697 parameters.
2025-02-25 10:53:15 (INFO): Loading dataset: lmdb
2025-02-25 10:53:15 (WARNING): Could not find dataset metadata.npz files in '[PosixPath('/data/wyx/OC20/is2re/10k/train/data.lmdb')]'
2025-02-25 10:53:15 (WARNING): Disabled BalancedBatchSampler because num_replicas=1.
2025-02-25 10:53:15 (WARNING): Failed to get data sizes, falling back to uniform partitioning. BalancedBatchSampler requires a dataset that has a metadata attributed with number of atoms.
2025-02-25 10:53:15 (INFO): rank: 0: Sampler created...
2025-02-25 10:53:15 (INFO): Created BalancedBatchSampler with sampler=<fairchem.core.common.data_parallel.StatefulDistributedSampler object at 0x7fab5042b5f0>, batch_size=64, drop_last=False
2025-02-25 10:53:15 (WARNING): Could not find dataset metadata.npz files in '[PosixPath('/data/wyx/OC20/is2re/all/val_id/data.lmdb')]'
2025-02-25 10:53:15 (WARNING): Disabled BalancedBatchSampler because num_replicas=1.
2025-02-25 10:53:15 (WARNING): Failed to get data sizes, falling back to uniform partitioning. BalancedBatchSampler requires a dataset that has a metadata attributed with number of atoms.
2025-02-25 10:53:15 (INFO): rank: 0: Sampler created...
2025-02-25 10:53:15 (INFO): Created BalancedBatchSampler with sampler=<fairchem.core.common.data_parallel.StatefulDistributedSampler object at 0x7fab5042ba40>, batch_size=64, drop_last=False
2025-02-25 10:53:15 (INFO): normalizers checkpoint for targets ['energy'] have been saved to: /home/wyx/AI4sci/fairchem/checkpoints/2025-02-25-10-52-48/normalizers.pt
2025-02-25 10:53:15 (INFO): Normalization values for output energy: mean=-1.525913953781128, rmsd=2.279365062713623.
Traceback (most recent call last):
File "/home/wyx/AI4sci/fairchem/main.py", line 8, in
main()
File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/_cli.py", line 135, in main
runner_wrapper(config)
File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/_cli.py", line 58, in runner_wrapper
Runner()(config)
File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/_cli.py", line 37, in call
with new_trainer_context(config=config) as ctx:
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/contextlib.py", line 137, in enter
return next(self.gen)
^^^^^^^^^^^^^^
File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/common/utils.py", line 1102, in new_trainer_context
trainer = trainer_cls(**trainer_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/trainers/ocp_trainer.py", line 109, in init
super().init(
File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/trainers/base_trainer.py", line 220, in init
self.load(inference_only)
File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/trainers/base_trainer.py", line 248, in load
self.load_loss()
File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/trainers/base_trainer.py", line 683, in load_loss
"coefficient" in loss[target]
AssertionError: 'coefficient' is not defined in the energy loss config {'fn': 'mae'}.
[rank0]: Traceback (most recent call last):
[rank0]: File "/home/wyx/AI4sci/fairchem/main.py", line 8, in
[rank0]: main()
[rank0]: File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/_cli.py", line 135, in main
[rank0]: runner_wrapper(config)
[rank0]: File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/_cli.py", line 58, in runner_wrapper
[rank0]: Runner()(config)
[rank0]: File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/_cli.py", line 37, in call
[rank0]: with new_trainer_context(config=config) as ctx:
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/contextlib.py", line 137, in enter
[rank0]: return next(self.gen)
[rank0]: ^^^^^^^^^^^^^^
[rank0]: File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/common/utils.py", line 1102, in new_trainer_context
[rank0]: trainer = trainer_cls(**trainer_config)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/trainers/ocp_trainer.py", line 109, in init
[rank0]: super().init(
[rank0]: File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/trainers/base_trainer.py", line 220, in init
[rank0]: self.load(inference_only)
[rank0]: File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/trainers/base_trainer.py", line 248, in load
[rank0]: self.load_loss()
[rank0]: File "/home/wyx/anaconda3/envs/fair-chem/lib/python3.12/site-packages/fairchem/core/trainers/base_trainer.py", line 683, in load_loss
[rank0]: "coefficient" in loss[target]
[rank0]: AssertionError: 'coefficient' is not defined in the energy loss config {'fn': 'mae'}.

How can I solve this problem?
Thank you!

[blbhc]

I also encountered this problem, and I solved it using the following method:

1.find the base.yml file in the folder of the dataset you are using, for example: Fairchem/configs/is2re/10k/base.yml

2.find the following code:

loss_functions:
- energy:
    coefficient: 4  # add this
    fn: mae

[misko]

Thank you @blbhc , that looks right to me! @yxwang1215 did this help resolve the issue?

[brunosamp-usp]

I am still getting the same error, even after i did the correction suggested by @blbhc . Is there any other way to fix this issue? I am using this to run:

python main.py --mode train --cpu --config-yml /home/brunoss/programs/fairchem/configs/s2ef/2M/equiformer_v2/equiformer_v2_N@12_L@[email protected]

The error:

Traceback (most recent call last):
File "/home/brunoss/programs/fairchem/main.py", line 8, in
main()
File "/home/brunoss/programs/fairchem/src/fairchem/core/_cli.py", line 129, in main
runner_wrapper(config)
File "/home/brunoss/programs/fairchem/src/fairchem/core/_cli.py", line 58, in runner_wrapper
Runner()(config)
File "/home/brunoss/programs/fairchem/src/fairchem/core/_cli.py", line 37, in call
with new_trainer_context(config=config) as ctx:
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/brunoss/miniconda3/lib/python3.12/contextlib.py", line 137, in enter
return next(self.gen)
^^^^^^^^^^^^^^
File "/home/brunoss/programs/fairchem/src/fairchem/core/common/utils.py", line 1103, in new_trainer_context
trainer = trainer_cls(**trainer_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/ocp_trainer.py", line 109, in init
super().init(
File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/base_trainer.py", line 220, in init
self.load(inference_only)
File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/base_trainer.py", line 246, in load
self.load_datasets()
File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/base_trainer.py", line 366, in load_datasets
self.train_sampler = self.get_sampler(
^^^^^^^^^^^^^^^^^
File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/base_trainer.py", line 314, in get_sampler
return BalancedBatchSampler(
^^^^^^^^^^^^^^^^^^^^^
File "/home/brunoss/programs/fairchem/src/fairchem/core/common/data_parallel.py", line 173, in init
raise error
File "/home/brunoss/programs/fairchem/src/fairchem/core/common/data_parallel.py", line 170, in init
dataset = _ensure_supported(dataset)
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/brunoss/programs/fairchem/src/fairchem/core/common/data_parallel.py", line 113, in _ensure_supported
raise UnsupportedDatasetError(
fairchem.core.datasets.base_dataset.UnsupportedDatasetError: BalancedBatchSampler requires a dataset that has a metadata attributed with number of atoms.
[rank0]: Traceback (most recent call last):
[rank0]: File "/home/brunoss/programs/fairchem/main.py", line 8, in
[rank0]: main()
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/_cli.py", line 129, in main
[rank0]: runner_wrapper(config)
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/_cli.py", line 58, in runner_wrapper
[rank0]: Runner()(config)
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/_cli.py", line 37, in call
[rank0]: with new_trainer_context(config=config) as ctx:
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/home/brunoss/miniconda3/lib/python3.12/contextlib.py", line 137, in enter
[rank0]: return next(self.gen)
[rank0]: ^^^^^^^^^^^^^^
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/common/utils.py", line 1103, in new_trainer_context
[rank0]: trainer = trainer_cls(**trainer_config)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/ocp_trainer.py", line 109, in init
[rank0]: super().init(
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/base_trainer.py", line 220, in init
[rank0]: self.load(inference_only)
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/base_trainer.py", line 246, in load
[rank0]: self.load_datasets()
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/base_trainer.py", line 366, in load_datasets
[rank0]: self.train_sampler = self.get_sampler(
[rank0]: ^^^^^^^^^^^^^^^^^
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/trainers/base_trainer.py", line 314, in get_sampler
[rank0]: return BalancedBatchSampler(
[rank0]: ^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/common/data_parallel.py", line 173, in init
[rank0]: raise error
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/common/data_parallel.py", line 170, in init
[rank0]: dataset = _ensure_supported(dataset)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/home/brunoss/programs/fairchem/src/fairchem/core/common/data_parallel.py", line 113, in _ensure_supported
[rank0]: raise UnsupportedDatasetError(
[rank0]: fairchem.core.datasets.base_dataset.UnsupportedDatasetError: BalancedBatchSampler requires a dataset that has a metadata attributed with number of atoms.

[blbhc]

@brunosamp-usp Perhaps you could try modifying the equiformer_v2_N@12_L@[email protected] file as follows:

optim:
  batch_size:                   4         # 6
  eval_batch_size:              4         # 6
  load_balancing: neighbors     # Modify this line
  num_workers: 8
  lr_initial:                   0.0004    # [0.0002, 0.0004], eSCN uses 0.0008 for batch size 96

I'm not sure if this solution will solve your problem because a 2M dataset is too big for me, so I didn't reproduce your problem, but you can try it.

By the way, this problem you're experiencing has been brought up before, so maybe you can check it out: #868