Last Week’s Work Review#
Recall our Phase 1 visual module training scheme
Our Phase 2 language module training scheme
the testing accuracy could achieve $99.0%$ But I doubt about the true performance of this model.
- the model can achieve high accuracy by just memorising the current 3D grid value as long as the incremental change is small.
- which is true for this dataset because one instruction is often paired with a small change in the environment
- I should change the current training scheme to prevent model from taking shortcuts
- local addition
- one step (no intermediate states)
- local addition
After completing language module, our next step is to build PDDL model to bridge the gap between current state to the target state described by the instruction.
You need password to access to the content, go to Slack *#phdsukai to find more.
Part of this article is encrypted with password:
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