fix: Allow full model compilation with collection inputs (`input_signature`) by gs-olive · Pull Request #1656 · pytorch/TensorRT
gs-olive
marked this pull request as ready for review
- Allow users to specify full model compilation when using `input_signature`, which allows for complex collection-based inputs - Enable "psuedo-partitioning" phase for input collections as well as output collections - Update `OutputIsCollection` to include dictionary outputs, and add function `InputIsCollection` to detect collection-based inputs during graph compilation - Remove automatic fallback for collection pack/unpack operations when using `input_signature` argument - Add collections tests to ensure full compilation is respected for input and output collections
gs-olive
deleted the
input_signature_full_compilation
branch
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters