Issues emerged when followed the tutorial

[julikoapp]

Hello,

We followed the official tutorial (create an environment and run pip install "syntheseus[all]" to install synthesus on Mac M1. The first problem was that the environment_full.yaml was given only for the GPUs cases, so we adapted it the following way:

channels:
- defaults
- conda-forge
- pytorch
- dglteam # LocalRetro, RetroKNN
- pyg # RetroKNN
dependencies:
- python==3.8.20 # MHNreact

Installing the needed packages additionally when needed to run the models:

torch                    2.1.2
rdkit                    2024.3.5
pytorch-lightning        1.9.4
dgl                      0.9.1
...

When the problems with package dependencies were solved, there was the following error when we tried to run the tutorial code:

Traceback (most recent call last):
  File "test.py", line 39, in <module>
    evaluate_with_syntheseus("C1CCCCC1")
  File "test.py", line 31, in evaluate_with_syntheseus
    model = LocalRetroModel()
  File "/Users/Ulia/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/local_retro.py", line 32, in __init__
    super().__init__(*args, **kwargs)
  File "/Users/Ulia/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/base.py", line 23, in __init__
    self.model_dir = Path(model_dir or self.get_default_model_dir())
  File "/Users/Ulia/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/base.py", line 27, in get_default_model_dir
    model_dir = get_default_model_dir_from_cache(self.name, is_forward=self.is_forward())
  File "/Users/Ulia/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/base.py", line 39, in name
    return self.__class__.__name__.removesuffix("Model")
AttributeError: 'str' object has no attribute 'removesuffix'

To solve this in "/Users/Ulia/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/base.py" the lines

@property
    def name(self) -> str:
        return self.__class__.__name__.removesuffix("Model")

were changed to

@property
    def name(self) -> str:
        def remove_suffix(src, suffix):
            if suffix and src.endswith(suffix):
                return src[:-len(suffix)]
            return src[:]
        return remove_suffix(self.__class__.__name__, "Model")

Also, a few errors emerged when running the models:

1. RootAlignedModel

"""
Traceback (most recent call last):
  File "/Users/Ulia/miniconda3/envs/syntheseus_ver_2/lib/python3.8/multiprocessing/pool.py", line 125, in worker
    result = (True, func(*args, **kwds))
  File "/Users/Ulia/miniconda3/envs/syntheseus_ver_2/lib/python3.8/multiprocessing/pool.py", line 48, in mapstar
    return list(map(*args))
  File "/Users/Ulia/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/root_aligned/score.py", line 23, in canonicalize_smiles_clear_map
    mol = Chem.MolFromSmiles(smiles,sanitize=not opt.synthon)
NameError: name 'opt' is not defined
"""

The above exception was the direct cause of the following exception:

NameError                                 Traceback (most recent call last)
Cell In[17], [line 13](vscode-notebook-cell:?execution_count=17&line=13)
     [11](vscode-notebook-cell:?execution_count=17&line=11) test_mol = Molecule("Cc1ccc(-c2ccc(C)cc2)cc1")
     [12](vscode-notebook-cell:?execution_count=17&line=12) model = RootAlignedModel()
---> [13](vscode-notebook-cell:?execution_count=17&line=13) [results] = model([test_mol], num_results=5)
     [14](vscode-notebook-cell:?execution_count=17&line=14) print_results(results)

File ~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/interface/models.py:110, in ReactionModel.__call__(self, inputs, num_results)
    [108](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/interface/models.py:108) inputs_not_in_cache = list({inp for inp in inputs if (inp, num_results) not in self._cache})
...
    [769](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/multiprocessing/pool.py:769)     return self._value
    [770](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/multiprocessing/pool.py:770) else:
--> [771](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/multiprocessing/pool.py:771)     raise self._value

NameError: name 'opt' is not defined

2. ChemformerModel

TypeError                                 Traceback (most recent call last)
Cell In[19], [line 13](vscode-notebook-cell:?execution_count=19&line=13)
     [11](vscode-notebook-cell:?execution_count=19&line=11) test_mol = Molecule("Cc1ccc(-c2ccc(C)cc2)cc1")
     [12](vscode-notebook-cell:?execution_count=19&line=12) model = ChemformerModel()
---> [13](vscode-notebook-cell:?execution_count=19&line=13) [results] = model([test_mol], num_results=5)
     [14](vscode-notebook-cell:?execution_count=19&line=14) print_results(results)

File ~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/interface/models.py:110, in ReactionModel.__call__(self, inputs, num_results)
    [108](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/interface/models.py:108) inputs_not_in_cache = list({inp for inp in inputs if (inp, num_results) not in self._cache})
    [109](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/interface/models.py:109) if len(inputs_not_in_cache) > 0:
--> [110](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/interface/models.py:110)     new_rxns = self._get_reactions(inputs=inputs_not_in_cache, num_results=num_results)
    [111](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/interface/models.py:111)     assert len(new_rxns) == len(inputs_not_in_cache)
    [112](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/interface/models.py:112)     for inp, rxns in zip(inputs_not_in_cache, new_rxns):

File ~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:130, in ChemformerModel._get_reactions(self, inputs, num_results)
    [127](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:127) with torch.no_grad(), warnings.catch_warnings():
    [128](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:128)     warnings.filterwarnings("ignore", message="__floordiv__ is deprecated")
--> [130](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:130)     smiles_batch, batch_log_likelihoods = self.model.sample_molecules(
    [131](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:131)         device_batch, sampling_alg="beam"
    [132](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:132)     )
    [134](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:134) # Choose processing function (controls reaction types).
    [135](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:135) # `type: ignore[assignment]` statements are because link between is_forward/is_backward
    [136](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:136) # and [InputType, ReactionType] is not visible to mypy.
    [137](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/syntheseus/reaction_prediction/inference/chemformer.py:137) if self.is_forward():
...
-> [1527](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/torch/nn/modules/module.py:1527)     return forward_call(*args, **kwargs)
   [1529](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/torch/nn/modules/module.py:1529) try:
   [1530](https://file+.vscode-resource.vscode-cdn.net/Users/Ulia/Desktop/aalto_summer_2025/code/retrosynthesis-dataset/~/miniconda3/envs/syntheseus_ver_2/lib/python3.8/site-packages/torch/nn/modules/module.py:1530)     result = None

TypeError: forward() got an unexpected keyword argument 'is_causal'

I didn't check all models for now. But an additional environment.yaml file for the CPU environments would be helpful.

[mrwnmsr]

Thanks you @julikoapp for these observations, so far we haven't tested on apple chips.
We will look into it.