Modifications for llama3.1-8b edge

script/app-mlperf-inference-mlcommons-python/meta.yaml

@@ -1449,10 +1449,9 @@ variations:
     env:
       MLC_GENERIC_PYTHON_PIP_EXTRA_FIND_LINKS_URL: "https://data.pyg.org/whl/torch-<<<MLC_TORCH_VERSION>>>+cpu.html"
 
-  llama3_1-8b:
-    group: models
-    env:
-      MLC_MODEL: llama3_1-8b
+  llama3_1-8b_:
+    group:
+      models
     deps:
       - tags: get,generic-python-lib,_package.transformers
       - tags: get,generic-python-lib,_package.sentencepiece
@@ -1469,27 +1468,35 @@ variations:
         names:
           - llama3_1-8b
           - llama3-8b
-      ## CNNDM for Llama3 8B model - edge
-      - tags: get,dataset,cnndm,_validation,_edge,_llama3,_mlc,_r2_downloader
-        names:
-          - cnndm-llama3-edge
-        enable_if_env:
-          MLC_MLPERF_SUBMISSION_SYSTEM_TYPE: 
-            - edge
-        skip_if_env:
-          MLC_RUN_STATE_DOCKER:
-          - "yes"
+
+  llama3_1-8b:
+    base:
+    - llama3_1-8b_
+    env:
+      MLC_MODEL: llama3_1-8b
+    deps:
       ## CNNDM for Llama3 8B model - datacenter
       - tags: get,dataset,cnndm,_validation,_datacenter,_llama3,_mlc,_r2_downloader
         names:
           - cnndm-llama3-datacenter
-        enable_if_env:
-          MLC_MLPERF_SUBMISSION_SYSTEM_TYPE: 
-            - datacenter
         skip_if_env:
           MLC_RUN_STATE_DOCKER:
           - "yes"
 
+  llama3_1-8b-edge:
+    base:
+    - llama3_1-8b_
+    env:
+      MLC_MODEL: llama3_1-8b-edge
+    deps:
+      ## CNNDM for Llama3 8B model - edge
+      - tags: get,dataset,cnndm,_validation,_edge,_llama3,_mlc,_r2_downloader
+        names:
+          - cnndm-llama3-edge
+        skip_if_env:
+          MLC_RUN_STATE_DOCKER:
+          - "yes"
+
   whisper:
     group: models
     env:

script/app-mlperf-inference/meta.yaml

@@ -955,10 +955,31 @@ variations:
       - mlperf-accuracy-script
       - waymo-accuracy-script
       tags: run,accuracy,mlperf,_waymo
-    
-  llama3_1-8b:
+
+  llama3_1-8b_:
     group:
       model
+    adr:
+      numpy:
+        version_max: "1.26.999"
+        version_max_usable: "1.26.4"
+    docker:
+      base_image: ubuntu:24.04
+      mounts:
+        - "${{ MLC_ML_MODEL_LLAMA3_CHECKPOINT_PATH }}:${{ MLC_ML_MODEL_LLAMA3_CHECKPOINT_PATH }}" 
+        - "${{ MLC_DATASET_CNNDM_EVAL_PATH }}:${{ MLC_DATASET_CNNDM_EVAL_PATH }}"
+      deps:
+      - tags: get,ml-model,llama3,_mlc,_8b,_r2_downloader
+        enable_if_env:
+          MLC_USE_ML_MODEL_FROM_HOST:
+          - 'yes'
+        names:
+          - llama3_1-8b
+          - llama3-8b
+
+  llama3_1-8b:
+    base:
+    - llama3_1-8b_
     add_deps_recursive:
       mlperf-inference-implementation:
         tags: _llama3_1-8b
@@ -972,42 +993,50 @@ variations:
         - all
         MLC_MLPERF_ACCURACY_RESULTS_DIR:
         - 'on'
-        MLC_MLPERF_SUBMISSION_SYSTEM_TYPE: 
-        - edge
       skip_if_env:
         MLC_MLPERF_IMPLEMENTATION:
         - nvidia
       names:
       - mlperf-accuracy-script
       - llama3_1-8b-accuracy-script
-      tags: run,accuracy,mlperf,_cnndm_llama_3,_edge
+      tags: run,accuracy,mlperf,_cnndm_llama_3,_datacenter
+    docker:
+      deps:
+      ## CNNDM for Llama3 8B model - datacenter
+      - tags: get,dataset,cnndm,_validation,_datacenter,_llama3,_mlc,_r2_downloader
+        names:
+          - cnndm-llama3-datacenter
+        enable_if_env:
+          MLC_MLPERF_SUBMISSION_SYSTEM_TYPE: 
+            - datacenter
+          MLC_USE_DATASET_FROM_HOST:
+          - 'yes'
+
+  llama3.1-8b-edge:
+    base:
+    - llama3_1-8b_
+    add_deps_recursive:
+      mlperf-inference-implementation:
+        tags: _llama3_1-8b-edge
+    env:
+      MLC_MODEL:
+        llama3_1-8b-edge
+    posthook_deps:
     - enable_if_env:
         MLC_MLPERF_LOADGEN_MODE:
         - accuracy
         - all
         MLC_MLPERF_ACCURACY_RESULTS_DIR:
         - 'on'
-        MLC_MLPERF_SUBMISSION_SYSTEM_TYPE: 
-        - datacenter
       skip_if_env:
         MLC_MLPERF_IMPLEMENTATION:
         - nvidia
       names:
       - mlperf-accuracy-script
       - llama3_1-8b-accuracy-script
-      tags: run,accuracy,mlperf,_cnndm_llama_3,_datacenter
+      tags: run,accuracy,mlperf,_cnndm_llama_3,_edge
     docker:
-      mounts:
-        - "${{ MLC_ML_MODEL_LLAMA3_CHECKPOINT_PATH }}:${{ MLC_ML_MODEL_LLAMA3_CHECKPOINT_PATH }}" 
-        - "${{ MLC_DATASET_CNNDM_EVAL_PATH }}:${{ MLC_DATASET_CNNDM_EVAL_PATH }}"
       deps:
-      - tags: get,ml-model,llama3,_mlc,_8b,_r2_downloader
-        enable_if_env:
-          MLC_USE_ML_MODEL_FROM_HOST:
-          - 'yes'
-        names:
-          - llama3_1-8b
-          - llama3-8b
       ## CNNDM for Llama3 8B model - edge
       - tags: get,dataset,cnndm,_validation,_edge,_llama3,_mlc,_r2_downloader
         names:
@@ -1017,15 +1046,6 @@ variations:
             - edge
           MLC_USE_DATASET_FROM_HOST:
           - 'yes'
-      ## CNNDM for Llama3 8B model - datacenter
-      - tags: get,dataset,cnndm,_validation,_datacenter,_llama3,_mlc,_r2_downloader
-        names:
-          - cnndm-llama3-datacenter
-        enable_if_env:
-          MLC_MLPERF_SUBMISSION_SYSTEM_TYPE: 
-            - datacenter
-          MLC_USE_DATASET_FROM_HOST:
-          - 'yes'
 
   whisper:
     group:

script/get-dataset-cnndm/meta.yaml

@@ -153,7 +153,7 @@ variations:
       MLC_DOWNLOAD_URL: mlc-inference:mlcommons-inference-wg-public/llama3.1_8b/<<<MLC_DATASET_CNNDM_FILENAME>>>
       MLC_DOWNLOAD_FINAL_ENV_NAME: MLC_DATASET_CNNDM_EVAL_PATH
       MLC_EXTRACT_FINAL_ENV_NAME: MLC_DATASET_CNNDM_EVAL_PATH
-  validation,datacenter,llama3,mlc,r2_downlaoder:
+  validation,datacenter,llama3,mlc,r2_downloader:
     adr:
       dae:
         extra_cache_tags: cnndm,dataset,llama3,val,datacenter
@@ -188,4 +188,4 @@ tests:
       - validation,edge,r2_downloader,llama3,mlc,dry-run
       - validation,datacenter,r2_downloader,llama3,mlc,dry-run
       - calibration,r2_downloader,llama3,mlc,dry-run
-      # - calibration,rclone,llama3,mlc,dry-run
+      # - calibration,rclone,llama3,mlc,dry-run