Mike Iovine

In my project, I have to test for equality of `std::pair`s. In other testing frameworks, this is very easy. For instance, in GoogleTest, this works: `ASSERT_EQ(std::make_pair(...), std::make_pair(...))`. This task turns...

feat: Add initial EAGLE-3 implementation

20

Add an initial implementation of EAGLE3. The algorithm is not fully implemented yet. Specifically: we are predicting **single extension sequences only** with this MR, not the fully token trees introduced...

[nvbug/5285881][fix] Fix chunked prefill + overlap scheduler

3

# [nvbug/5285881][fix] Fix chunked prefill + overlap scheduler ## Description The `token_idx` needed to be advanced when the decoder processes chunked requests during overlap scheduling ## Test Coverage Unwaived existing...

[None][chore] Weekly mass integration of release/1.1

50

## Description PRs explicitly excluded in this round: * https://github.com/NVIDIA/TensorRT-LLM/pull/8621: DLFW upgrade, already applied in main branch. * https://github.com/NVIDIA/TensorRT-LLM/pull/8877: Another DLFW related upgrade. * https://github.com/NVIDIA/TensorRT-LLM/pull/8860: CI change intended for release...