# TypeError: Object of type int32 is not JSON serializable **URL:** \ **Category:** Uncategorized\ **Created:** [January 6, 2023, 6:54pm UTC](https://discuss.pytorch.org/t/typeerror-object-of-type-int32-is-not-json-serializable/169762 "2023-01-06T18:54:00Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![mattpats](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/m/dc4da7/32.png) [@mattpats](https://discuss.pytorch.org/u/mattpats)\ **Post date:** [January 6, 2023, 6:54pm UTC](https://discuss.pytorch.org/t/typeerror-object-of-type-int32-is-not-json-serializable/169762/1 "2023-01-06T18:54:00Z")
I am training a TabNetClassifier. My code is largely borrowed from: ‘[tabnet/census\_example.ipynb at develop · dreamquark-ai/tabnet · GitHub](https://github.com/dreamquark-ai/tabnet/blob/develop/census_example.ipynb)’ Everything is working fine until I try to save the model. When trying to save the model, I get the error: ‘TypeError: Object of type int32 is not JSON serializable’ More details below: ```auto from pytorch_tabnet.tab_model import TabNetClassifier tabnet_params = {"cat_idxs": cat_idxs, "cat_dims": cat_dims, "cat_emb_dim": 2, "optimizer_fn": torch.optim.Adam, "optimizer_params": dict(lr=2e-2), "scheduler_params": {"step_size": 50, "gamma": 0.9}, "scheduler_fn": torch.optim.lr_scheduler.StepLR, "mask_type": 'entmax'} clf = TabNetClassifier(**tabnet_params) clf.fit( X_train=X_train, y_train=y_train, eval_set=[(X_train, y_train), (X_valid, y_valid)], eval_name=['train', 'valid'], eval_metric=['auc'], max_epochs=50, patience=20, batch_size=1024, virtual_batch_size=128, num_workers=0, weights=1, drop_last=False) ``` My model trains fine and I can perform inference with it. However, when I try to save it: ```auto saving_path_name = "./tabnet_model_test_1" saved_filepath = clf.save_model(saving_path_name) >>> TypeError: Object of type int32 is not JSON serializable ``` ---
**Author:** ![rinkujadhav2013](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/rinkujadhav2013/32/54900_2.png) [@rinkujadhav2013](https://discuss.pytorch.org/u/rinkujadhav2013)\ **Post date:** [January 7, 2023, 1:39am UTC](https://discuss.pytorch.org/t/typeerror-object-of-type-int32-is-not-json-serializable/169762/2 "2023-01-07T01:39:43Z")
Maybe have a look here: > > > \`numpy.bool\_\` types are not being correctly serialized to json. > > \*\*What is the… current behavior?\*\* > The \`ComplexEncoder\` class (\[here\](https://github.com/dreamquark-ai/tabnet/blob/4fa545da50796f0d16f49d0cb476d5a30c2a27c1/pytorch\_tabnet/utils.py#L336)) does not handle \`numpy.bool\_\` which is not JSON serializable. This raises a TypeError when saving certain models. > > \*\*If the current behavior is a bug, please provide the steps to reproduce.\*\* > \`\`\`python > model = TabNetClassifier(...) > model.fit(...) # training data and model parameters contain values of type numpy.bool\_ > model.save\_model('path/to/model') > \`\`\` > > \*\*Expected behavior\*\* > \`numpy.bool\_\` should be cast to python's \`bool\` before being serialized to JSON. Here is my suggested fix. Please let me know if this is acceptable for a PR: > > \`\`\`python > class ComplexEncoder(json.JSONEncoder): > def default(self, obj): > if isinstance(obj, np.int64): > return int(obj) > if isinstance(obj, np.bool\_): > return bool(obj) > # Let the base class default method raise the TypeError > return json.JSONEncoder.default(self, obj) > \`\`\` > > \*\*Other relevant information:\*\* > poetry version: "poetry-core\>=1.0.0" > python version: "^3.9" > Operating System: "Linux Kernel 5.18.14-arch1-1" > Additional tools: > CUDA Version: 11.7 > Driver Version: 515.57 > > \*\*Additional context\*\* > > Here's a stacktrace: > \`\`\`bash > File ".venv/lib/python3.10/site-packages/pytorch\_tabnet/abstract\_model.py", line 375, in save\_model > json.dump(saved\_params, f, cls=ComplexEncoder) > File "/usr/lib/python3.10/json/\_\_init\_\_.py", line 179, in dump > for chunk in iterable: > File "/usr/lib/python3.10/json/encoder.py", line 431, in \_iterencode > yield from \_iterencode\_dict(o, \_current\_indent\_level) > File "/usr/lib/python3.10/json/encoder.py", line 405, in \_iterencode\_dict > yield from chunks > File "/usr/lib/python3.10/json/encoder.py", line 405, in \_iterencode\_dict > yield from chunks > File "/usr/lib/python3.10/json/encoder.py", line 405, in \_iterencode\_dict > yield from chunks > File "/usr/lib/python3.10/json/encoder.py", line 438, in \_iterencode > o = \_default(o) > File ".venv/lib/python3.10/site-packages/pytorch\_tabnet/utils.py", line 339, in default > return json.JSONEncoder.default(self, obj) > File "/usr/lib/python3.10/json/encoder.py", line 179, in default > raise TypeError(f'Object of type {o.\_\_class\_\_.\_\_name\_\_} ' > TypeError: Object of type bool\_ is not JSON serializable > \`\`\` > > I ran into this when trying tabnet in a \[kaggle competition\](https://www.kaggle.com/competitions/amex-default-prediction). If you need to, you can look \[here in my code\](https://github.com/nishaq503/amex/blob/b4dabc441581a6f29318fcbd2a83ec5376bc1f5a/amex/models/tabnet/training.py#L121) where the error happens. and here: > ---
**Author:** ![Magill\_Anway](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/magill_anway/32/79999_2.png) [@Magill\_Anway](https://discuss.pytorch.org/u/Magill_Anway)\ **Post date:** [September 5, 2026, 9:26am UTC](https://discuss.pytorch.org/t/typeerror-object-of-type-int32-is-not-json-serializable/169762/3 "2026-09-05T09:26:03Z")
The failing value is likely a NumPy scalar (for example numpy.int32), which the standard json encoder does not know. Convert NumPy scalars at the boundary with int(value) or value.item(); use .tolist() for NumPy arrays, and .item()/.tolist() for scalar or tensor values. For nested params, recursively normalize dicts and lists before json.dump. To quickly inspect the normalized output without uploading it, this browser-local JSON formatter is useful: [JSON Formatter | Free Online Tool | Lizely](https://www.lizecheng.net/dev/json-formatter/?utm_source=discuss.pytorch.org&utm_medium=forum&utm_campaign=lizely_track1&utm_content=json-formatter)