From 51831b435bba226df5c4e3d2fbe1daed0008f289 Mon Sep 17 00:00:00 2001 From: anandhu-eng Date: Fri, 25 Jul 2025 11:01:19 +0530 Subject: [PATCH 01/11] Modifications for llama3 edge --- .../meta.yaml | 41 ++++++----- script/app-mlperf-inference/meta.yaml | 69 ++++++++++++------- 2 files changed, 67 insertions(+), 43 deletions(-) diff --git a/script/app-mlperf-inference-mlcommons-python/meta.yaml b/script/app-mlperf-inference-mlcommons-python/meta.yaml index c5ebee492..205632003 100644 --- a/script/app-mlperf-inference-mlcommons-python/meta.yaml +++ b/script/app-mlperf-inference-mlcommons-python/meta.yaml @@ -1447,10 +1447,9 @@ variations: env: MLC_GENERIC_PYTHON_PIP_EXTRA_FIND_LINKS_URL: "https://data.pyg.org/whl/torch-<<>>+cpu.html" - llama3_1-8b: - group: models - env: - MLC_MODEL: llama3_1-8b + llama3_1-8_: + group: + models deps: - tags: get,generic-python-lib,_package.transformers - tags: get,generic-python-lib,_package.sentencepiece @@ -1467,27 +1466,35 @@ 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_1-8_ + 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_1-8_ + 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: diff --git a/script/app-mlperf-inference/meta.yaml b/script/app-mlperf-inference/meta.yaml index b1d7ef38e..08cefd158 100644 --- a/script/app-mlperf-inference/meta.yaml +++ b/script/app-mlperf-inference/meta.yaml @@ -955,10 +955,26 @@ variations: - mlperf-accuracy-script - waymo-accuracy-script tags: run,accuracy,mlperf,_waymo - - llama3_1-8b: + + llama3_1-8_: group: model + 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 + + llama3_1-8b: + base: + - llama3_1-8_ add_deps_recursive: mlperf-inference-implementation: tags: _llama3_1-8b @@ -972,15 +988,35 @@ 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_1-8_ + 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 @@ -988,26 +1024,16 @@ variations: MLC_MLPERF_ACCURACY_RESULTS_DIR: - 'on' MLC_MLPERF_SUBMISSION_SYSTEM_TYPE: - - datacenter + - edge 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 +1043,6 @@ 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: From 570688246b709d1078c3ca4a0d523ab6023fe212 Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 18:35:00 +0530 Subject: [PATCH 02/11] fix typo --- script/get-dataset-cnndm/meta.yaml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/script/get-dataset-cnndm/meta.yaml b/script/get-dataset-cnndm/meta.yaml index 1e891215c..4d38491f8 100644 --- a/script/get-dataset-cnndm/meta.yaml +++ b/script/get-dataset-cnndm/meta.yaml @@ -153,7 +153,7 @@ variations: MLC_DOWNLOAD_URL: mlc-inference:mlcommons-inference-wg-public/llama3.1_8b/<<>> 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 \ No newline at end of file + # - calibration,rclone,llama3,mlc,dry-run From 0d3ecccd386ca4e6811c9538595becda9aa4432b Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 19:07:56 +0530 Subject: [PATCH 03/11] update base image for llama3.1 8b --- script/app-mlperf-inference/meta.yaml | 1 + 1 file changed, 1 insertion(+) diff --git a/script/app-mlperf-inference/meta.yaml b/script/app-mlperf-inference/meta.yaml index 08cefd158..359976835 100644 --- a/script/app-mlperf-inference/meta.yaml +++ b/script/app-mlperf-inference/meta.yaml @@ -963,6 +963,7 @@ variations: mounts: - "${{ MLC_ML_MODEL_LLAMA3_CHECKPOINT_PATH }}:${{ MLC_ML_MODEL_LLAMA3_CHECKPOINT_PATH }}" - "${{ MLC_DATASET_CNNDM_EVAL_PATH }}:${{ MLC_DATASET_CNNDM_EVAL_PATH }}" + base_image: nvcr.io/nvidia/pytorch:24.07-py3 deps: - tags: get,ml-model,llama3,_mlc,_8b,_r2_downloader enable_if_env: From e67ae34d4e7a98c8837a7800a4afa5b9ec54d5d3 Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 19:17:53 +0530 Subject: [PATCH 04/11] Update meta.yaml --- script/app-mlperf-inference/meta.yaml | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/script/app-mlperf-inference/meta.yaml b/script/app-mlperf-inference/meta.yaml index 359976835..bc269aba2 100644 --- a/script/app-mlperf-inference/meta.yaml +++ b/script/app-mlperf-inference/meta.yaml @@ -963,7 +963,6 @@ variations: mounts: - "${{ MLC_ML_MODEL_LLAMA3_CHECKPOINT_PATH }}:${{ MLC_ML_MODEL_LLAMA3_CHECKPOINT_PATH }}" - "${{ MLC_DATASET_CNNDM_EVAL_PATH }}:${{ MLC_DATASET_CNNDM_EVAL_PATH }}" - base_image: nvcr.io/nvidia/pytorch:24.07-py3 deps: - tags: get,ml-model,llama3,_mlc,_8b,_r2_downloader enable_if_env: @@ -973,6 +972,14 @@ variations: - llama3_1-8b - llama3-8b + llama3_1-8_,cpu: + docker: + base_image: ubuntu;24.04 + + llama3_1-8_,cuda: + docker: + base_image: nvcr.io/nvidia/pytorch:24.07-py3 + llama3_1-8b: base: - llama3_1-8_ From 96b60f56fdb32371343a5d33f414a75c9181ffff Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 19:19:13 +0530 Subject: [PATCH 05/11] fix typo --- script/app-mlperf-inference/meta.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/script/app-mlperf-inference/meta.yaml b/script/app-mlperf-inference/meta.yaml index bc269aba2..0fee74cd2 100644 --- a/script/app-mlperf-inference/meta.yaml +++ b/script/app-mlperf-inference/meta.yaml @@ -974,7 +974,7 @@ variations: llama3_1-8_,cpu: docker: - base_image: ubuntu;24.04 + base_image: ubuntu:24.04 llama3_1-8_,cuda: docker: From f04cb8c6cec88ddd95d48016b2785a6f2fab5d21 Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 20:20:18 +0530 Subject: [PATCH 06/11] Update meta.yaml --- script/app-mlperf-inference/meta.yaml | 9 +-------- 1 file changed, 1 insertion(+), 8 deletions(-) diff --git a/script/app-mlperf-inference/meta.yaml b/script/app-mlperf-inference/meta.yaml index 0fee74cd2..af8799ea3 100644 --- a/script/app-mlperf-inference/meta.yaml +++ b/script/app-mlperf-inference/meta.yaml @@ -960,6 +960,7 @@ variations: group: model 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 }}" @@ -971,14 +972,6 @@ variations: names: - llama3_1-8b - llama3-8b - - llama3_1-8_,cpu: - docker: - base_image: ubuntu:24.04 - - llama3_1-8_,cuda: - docker: - base_image: nvcr.io/nvidia/pytorch:24.07-py3 llama3_1-8b: base: From db73f7846b175cf2e681a4e422d2a36f2c62f84f Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 20:34:17 +0530 Subject: [PATCH 07/11] fix numpy version --- script/app-mlperf-inference/meta.yaml | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/script/app-mlperf-inference/meta.yaml b/script/app-mlperf-inference/meta.yaml index af8799ea3..aa8abb178 100644 --- a/script/app-mlperf-inference/meta.yaml +++ b/script/app-mlperf-inference/meta.yaml @@ -959,6 +959,12 @@ variations: llama3_1-8_: group: model + deps: + - names: + - numpy + tags: get,generic-python-lib,_package.numpy + version_max: "1.26.999" + version_max_usable: "1.26.4" docker: base_image: ubuntu:24.04 mounts: From 0e0d7c9d9e976c21843ffb9f4da00797178da95c Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 20:56:53 +0530 Subject: [PATCH 08/11] Update meta.yaml --- script/app-mlperf-inference/meta.yaml | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) diff --git a/script/app-mlperf-inference/meta.yaml b/script/app-mlperf-inference/meta.yaml index aa8abb178..a5c4b14c9 100644 --- a/script/app-mlperf-inference/meta.yaml +++ b/script/app-mlperf-inference/meta.yaml @@ -959,12 +959,10 @@ variations: llama3_1-8_: group: model - deps: - - names: - - numpy - tags: get,generic-python-lib,_package.numpy - version_max: "1.26.999" - version_max_usable: "1.26.4" + adr: + numpy: + version_max: "1.26.999" + version_max_usable: "1.26.4" docker: base_image: ubuntu:24.04 mounts: From 48ac7bad8102bed0250a2a7dc1ec5d366b098dd8 Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 21:23:53 +0530 Subject: [PATCH 09/11] clean code --- script/app-mlperf-inference/meta.yaml | 2 -- 1 file changed, 2 deletions(-) diff --git a/script/app-mlperf-inference/meta.yaml b/script/app-mlperf-inference/meta.yaml index a5c4b14c9..053e320af 100644 --- a/script/app-mlperf-inference/meta.yaml +++ b/script/app-mlperf-inference/meta.yaml @@ -1028,8 +1028,6 @@ variations: - all MLC_MLPERF_ACCURACY_RESULTS_DIR: - 'on' - MLC_MLPERF_SUBMISSION_SYSTEM_TYPE: - - edge skip_if_env: MLC_MLPERF_IMPLEMENTATION: - nvidia From cff83b7f4d9c0bccf021b02da303f1feca888fdb Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 21:48:58 +0530 Subject: [PATCH 10/11] Update meta.yaml --- script/app-mlperf-inference/meta.yaml | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/script/app-mlperf-inference/meta.yaml b/script/app-mlperf-inference/meta.yaml index 053e320af..16be0570f 100644 --- a/script/app-mlperf-inference/meta.yaml +++ b/script/app-mlperf-inference/meta.yaml @@ -956,7 +956,7 @@ variations: - waymo-accuracy-script tags: run,accuracy,mlperf,_waymo - llama3_1-8_: + llama3_1-8b_: group: model adr: @@ -979,7 +979,7 @@ variations: llama3_1-8b: base: - - llama3_1-8_ + - llama3_1-8b_ add_deps_recursive: mlperf-inference-implementation: tags: _llama3_1-8b @@ -1014,7 +1014,7 @@ variations: llama3.1-8b-edge: base: - - llama3_1-8_ + - llama3_1-8b_ add_deps_recursive: mlperf-inference-implementation: tags: _llama3_1-8b-edge From 8fe786db822018b2d5366b297a8a8f896563beef Mon Sep 17 00:00:00 2001 From: ANANDHU S <71482562+anandhu-eng@users.noreply.github.com> Date: Fri, 25 Jul 2025 21:49:46 +0530 Subject: [PATCH 11/11] Update meta.yaml --- script/app-mlperf-inference-mlcommons-python/meta.yaml | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/script/app-mlperf-inference-mlcommons-python/meta.yaml b/script/app-mlperf-inference-mlcommons-python/meta.yaml index 0fd377da3..197d8ec94 100644 --- a/script/app-mlperf-inference-mlcommons-python/meta.yaml +++ b/script/app-mlperf-inference-mlcommons-python/meta.yaml @@ -1449,7 +1449,7 @@ variations: env: MLC_GENERIC_PYTHON_PIP_EXTRA_FIND_LINKS_URL: "https://data.pyg.org/whl/torch-<<>>+cpu.html" - llama3_1-8_: + llama3_1-8b_: group: models deps: @@ -1471,7 +1471,7 @@ variations: llama3_1-8b: base: - - llama3_1-8_ + - llama3_1-8b_ env: MLC_MODEL: llama3_1-8b deps: @@ -1485,7 +1485,7 @@ variations: llama3_1-8b-edge: base: - - llama3_1-8_ + - llama3_1-8b_ env: MLC_MODEL: llama3_1-8b-edge deps: