DecisionTransformerInterpretability
DecisionTransformerInterpretability copied to clipboard
Complete Embedding visualizations
- [x] 1. Dot product between each output action.
- [x] 2. Dot product between each input action.
- [ ] 3. Dot product between each time embedding.
- [x] 4. Dot product between each input state. 980*980. Create filtered version of this.
- [ ] 5. Eigenvalues of the state embedding? Label them?
Use Cosine Similarity not dot product.