This is a real problem. It’s not a bug in the sense of an implementation error, it’s just a failure of our current initialization and warmup procedure. See the following current thread for an explanation of what’s going on: Offset multiplier initialization - #44 by spinkney
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Offset multiplier initialization | 48 | 2766 | April 25, 2025 | |
| Few of questions on <offset= ... ,multiplier= ... > | 3 | 901 | December 11, 2020 | |
| NCP with `offset`/`multiplier` vs doing it "by hand" | 1 | 148 | August 4, 2025 | |
| Hierarchical MPT, divergent transitions | 2 | 399 | November 18, 2023 | |
| Partial Pooling Scale Parameter is Inf | 9 | 688 | March 9, 2021 |