[Web] ORT Training Doesn't Work for ResNet (Uncaught 51885088)

### Describe the issue

Attempting to load a ResNet model via `ort.TrainingSession` using `onnxruntime-web/training` leads to the following cryptic error:

```
Uncaught 51885088
```

### To reproduce

Code to reproduce can be found here: https://github.com/jaymody/ort-web-training-experiments/tree/resnet

```bash
git clone -b resnet https://github.com/jaymody/ort-web-training-experiments
cd ort-web-training-experiments
python -m pip install torch torchvision onnx onnxruntime-training
python create_onnx_files.py "resnet"
npm install
npm run dev
```

I've also added the option to export a simple 2-layer network via `python create_onnx_files.py "mlp"` and that works fine, so at least to my knowledge, seems to be an architecture thing?

The resnet model is loaded using `torch.hub`.

### Urgency

_No response_

### ONNX Runtime Installation

Released Package

### ONNX Runtime Version or Commit ID

1.17.1

### Execution Provider

'wasm'/'cpu' (WebAssembly CPU)