Merge pull request #52 from arjunsuresh/mlperf-inference

Restrict Tensorrtllm build to language models
mlcommons · Jun 6, 2024 · 2589200 · 2589200
2 parents 36c5663 + d6c18a3
commit 2589200
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 1 deletion.
diff --git a/script/app-mlperf-inference/_cm.yaml b/script/app-mlperf-inference/_cm.yaml
@@ -248,6 +248,8 @@ variations:
     docker:
       deps:
         - tags: get,ml-model,gptj,_nvidia,_fp8
+    env:
+      BUILD_TRTLLM: 1
 
   nvidia-original:
     docker:
@@ -284,6 +286,7 @@ variations:
       cnndm-accuracy-script:
         tags: _int32
     env:
+      BUILD_TRTLLM: 0
       CM_MLPERF_IMPLEMENTATION: nvidia
       CM_SQUAD_ACCURACY_DTYPE: float16
       CM_IMAGENET_ACCURACY_DTYPE: int32

diff --git a/script/build-mlperf-inference-server-nvidia/_cm.yaml b/script/build-mlperf-inference-server-nvidia/_cm.yaml
@@ -241,7 +241,7 @@ versions:
         version: r4.0
       nvidia-scratch-space:
         tags: _version.4_1
-    env:
+    default_env:
       BUILD_TRTLLM: 1
     deps:
       - tags: get,generic,sys-util,_nlohmann-json3-dev