Mismatch between FastAI prediction and Pytorch prediction

Here is the FastAI model:

TabularModel(
(embeds): ModuleList()
(emb_drop): Dropout(p=0.0, inplace=False)
(bn_cont): BatchNorm1d(3, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(layers): Sequential(
(0): LinBnDrop(
(0): Linear(in_features=3, out_features=200, bias=False)
(1): ReLU(inplace=True)
(2): BatchNorm1d(200, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(1): LinBnDrop(
(0): Linear(in_features=200, out_features=100, bias=False)
(1): ReLU(inplace=True)
(2): BatchNorm1d(100, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(2): LinBnDrop(
(0): Linear(in_features=100, out_features=1, bias=True)
)
)
)

See: Tabular model | fastai

It correctly makes a prediction. When I make a Pytorch prediction based on the FastAI model, as follows, the prediction is wrong (15 vs. 51):

with torch.no_grad():
modelFinal = learn.model.eval().to(“cpu”)
modelFinal (None ,torch_tensor.float())

The input tensor is:

tensor([[2.0000, 3.1000, 4.0000]], dtype=torch.float64)

Any ideas why the mismatch?

Resolved: it was due to the FastAI normalize config for props in:

dls = TabularDataLoaders.from_csv(props = [normalize])

I removed the normalize, and now the pytorch result matches FastAI.