The pytorch_model.bin from HuggingFace contains keys like:
txt_encoder.*
unified_encoder.*
decision_head.*
But the current code (navr1/models/) only defines:
NavR1Backbone (with language_encoder, vision_encoder, multimodal_fusion)
PolicyHead
The classes for txt_encoder, unified_encoder, etc. are not in the codebase.
Questions:
- Is there a matching code version for this checkpoint?
- Or is there an updated checkpoint that matches the current code?
Thanks!
The
pytorch_model.binfrom HuggingFace contains keys like:txt_encoder.*unified_encoder.*decision_head.*But the current code (
navr1/models/) only defines:NavR1Backbone(withlanguage_encoder,vision_encoder,multimodal_fusion)PolicyHeadThe classes for
txt_encoder,unified_encoder, etc. are not in the codebase.Questions:
Thanks!