TensorRT-LLM feat : reduce trt engine build time in testing

The default max tactic number to autotune is (when builder_config.trt_builder_config.max_num_tactics == -1 and trt will choose heuristic path): 100 for SM90 60 for other architectures This number may have slight impact on the e2e perf, but no impact on the functional tests and would save nearly 50% test time on related tests.

Mar 24 '25 08:03 peaceh-nv

/bot run

Mar 24 '25 08:03 peaceh-nv

PR_Github #261 [ run ] triggered by Bot

Mar 24 '25 08:03 niukuo

PR_Github #261 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #252 completed with status: 'FAILURE'

Mar 24 '25 09:03 niukuo

/bot run

Mar 24 '25 15:03 QiJune

PR_Github #317 [ run ] triggered by Bot

Mar 24 '25 16:03 niukuo

PR_Github #317 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #301 completed with status: 'FAILURE'

Mar 24 '25 18:03 niukuo

/bot run

Mar 25 '25 01:03 peaceh-nv

PR_Github #351 [ run ] triggered by Bot

Mar 25 '25 02:03 niukuo

PR_Github #351 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #322 completed with status: 'FAILURE'

Mar 25 '25 03:03 niukuo

/bot run

Mar 25 '25 04:03 peaceh-nv

PR_Github #374 [ run ] triggered by Bot

Mar 25 '25 04:03 niukuo

PR_Github #374 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #338 completed with status: 'FAILURE'

Mar 25 '25 06:03 niukuo

/bot run

Mar 25 '25 23:03 QiJune

PR_Github #476 [ run ] triggered by Bot

Mar 25 '25 23:03 niukuo

PR_Github #476 [ run ] completed with state SUCCESS /LLM/main/L0_MergeRequest_PR pipeline #409 completed with status: 'SUCCESS'

Mar 26 '25 01:03 niukuo

/bot reuse-pipeline

Mar 26 '25 03:03 peaceh-nv

PR_Github #506 [ reuse-pipeline ] triggered by Bot

Mar 26 '25 03:03 niukuo

PR_Github #506 [ reuse-pipeline ] completed with state SUCCESS Reusing PR_Github #476 for commit 2794aad

Mar 26 '25 03:03 niukuo