Update on the development branch

[kaiyux]

Hi,

The TensorRT-LLM team is pleased to announce that we have pushed an update to the development branch (and the Triton backend) this July 30, 2024.

This update includes:

• Features
  • Support Context Parallelism for DiT models, see “Context Parallel” section in examples/dit/README.md.
• Benchmark
  • A sink_token_len argument is added to gptManagerBenchmark.
  • Added encoder-decoder models support for gptManagerBenchmark.
• Infra
  • The dependent ModelOpt version has been upgraded to 0.15.0.

Thanks,
The TensorRT-LLM Engineering Team