kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-14 06:27:45 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

feat : reduce trt engine build time in testing (#3014)

Browse Source 
Signed-off-by: peaceh <103117813+peaceh-nv@users.noreply.github.com>
This commit is contained in:
peaceh-nv 2025-03-26 13:02:54 +08:00 committed by GitHub
parent 1ac0566a93
commit 5e272eef81
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
tests/unittest/attention/test_gpt_attention.py
View File

@ -754,7 +754,9 @@ class TestFunctional(unittest.TestCase):
precision=dtype,
int8=int8_trt_flag,
quant_mode=quant_mode)
# Reuce the TRT engine build time by setting the max allowed number of tactics in builder tactic profiling.
if builder_config.trt_builder_config.max_num_tactics == -1:
builder_config.trt_builder_config.max_num_tactics = 30
if session is None:
engine = builder.build_engine(net, builder_config)
session = tensorrt_llm.runtime.Session.from_serialized_engine(