From cee2ef6d86d3127e5bc5a542d96cee70bf15bf5f Mon Sep 17 00:00:00 2001 From: Frank Liu Date: Mon, 27 Sep 2021 19:52:47 -0700 Subject: [PATCH] Add python/tensorrt to benchmark Change-Id: Ib6b852fb08d3714bd30c3ada81b0c09118843ae7 --- extensions/benchmark/README.md | 4 +++- extensions/benchmark/build.gradle | 4 ++++ 2 files changed, 7 insertions(+), 1 deletion(-) diff --git a/extensions/benchmark/README.md b/extensions/benchmark/README.md index b00362db4fd..6190c4ae57a 100644 --- a/extensions/benchmark/README.md +++ b/extensions/benchmark/README.md @@ -22,8 +22,9 @@ djl-bench currently support benchmark the following type of models: - PaddlePaddle model - TFLite model - TensorRT model -- Neo DLR (TVM) model - XGBoost model +- Python script model +- Neo DLR (TVM) model You can build djl-bench from source if you need to benchmark fastText/BlazingText/Sentencepiece models. @@ -188,6 +189,7 @@ By default, the above script will use MXNet as the default Engine, but you can a -e TensorRT # TensorRT -e DLR # Neo DLR -e XGBoost # XGBoost +-e Python # Python script ``` ### Step 2: Identify the source of your model diff --git a/extensions/benchmark/build.gradle b/extensions/benchmark/build.gradle index a4d40837abf..5192f651468 100644 --- a/extensions/benchmark/build.gradle +++ b/extensions/benchmark/build.gradle @@ -33,6 +33,10 @@ dependencies { runtimeOnly "ai.djl.dlr:dlr-native-auto" runtimeOnly "ai.djl.ml.xgboost:xgboost" + + runtimeOnly "ai.djl.python:python" + + runtimeOnly "ai.djl.tensorrt:tensorrt" } else { implementation project(":model-zoo")