Kavyansh Tyagi

The model is just too large to fit your current setup cannot support DeepSeek-V3 671B. You’ll need 128 A100 GPUs (80GB each) for full inference. **Recommended Solutions** 1. Use a...

> I think this already works as expected. Am I missing something? The issue arises specifically when a `SparseTensor` is passed as `edge_index` into `GCNConv` with `cached=True`. After applying `gcn_norm`........

**Summary** LightningCLI's method add_lightning_class_args exposes a required parameter which is ignored when subclass_mode=False (the default), contradicting the documented behavior. The documentation claims that setting required=False should make providing the argument...

### Changes that might fix it : Initialize and maintain the membrane potentials outside the time step loop and pass mem explicitly through each time step instead of reinitializing it...