Refactor MNIST forward style and improve device selection

[amit-chaubey]

Summary

This PR improves readability in mnist/main.py and makes device selection explicit for CUDA/MPS fallback paths.

Changes

• Refactor Net.forward() to use direct functional composition:
  • x = F.relu(self.conv1(x))
  • x = F.relu(self.conv2(x))
  • x = F.relu(self.fc1(x))
  • return F.log_softmax(self.fc2(x), dim=1) directly
• Add concise type hints and short docstrings for forward, train, test, and main.
• Add _mps_available() helper and update device selection:
  • Prefer torch.accelerator.current_accelerator() when available
  • Fallback to torch.device("cuda"), then torch.device("mps"), else CPU
• Update use_accel gating so CUDA/MPS availability also enables accelerator code path.

Validation

• python -m py_compile mnist/main.py

[netlify]

✅ Deploy Preview for pytorch-examples-preview canceled.

Name	Link
🔨 Latest commit	5b84c52
🔍 Latest deploy log	https://app.netlify.com/projects/pytorch-examples-preview/deploys/69dcc4df7d2c1d0008b62b5b