TensorRT-LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and build TensorRT engines that contain state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT-LLM also contains components to create Python and C++ runtimes that execute those TensorRT engines.
max_output_len
withmax_seq_len
GptSession
Python bindingsgptManagerBenchmark
get_stats
andaget_stats
in HL Executor while using multi-gpuiterLatencyMilliSec
to stats and iteration logquantize.py
fails to export important data to config.jsonFAST_BUILD
ITensor::slice
to extreme casesNaN
appears in the result under the one shot all reduce strategyncclComm_t
asweak_ptr
and wrap it asshared_ptr
to avoid accidentally destroyed