cm4mlperf November 2024

What's Changed

• fixed several URLs (all tests passed) by @gfursin in #342
• Changes for supporting model and dataset download to host - Mixtral by @anandhu-eng in #346
• Support custom git clone branch in docker by @anandhu-eng in #343
• Merge from go, fixes #337 by @arjunsuresh in #348
• Includes cuda version to run suffix by @anandhu-eng in #354
• Fixes for const in script by @arjunsuresh in #355
• Fix docker image naming SCC24, extended CM script tests by @arjunsuresh in #356
• Fixes for MLPerf Inference Github Actions by @arjunsuresh in #362
• Fix typo in gh action by @arjunsuresh in #363
• Fix CUDA num_devices by @arjunsuresh in #365
• Support cleaning of Nvidia SDXL model by @arjunsuresh in #366
• Improvements to Nvidia MLPerf interface by @arjunsuresh in #367
• Fixes to pull changes for Nvidia implementation by @arjunsuresh in #369
• Improvements to MLPerf inference final report generation by @arjunsuresh in #371
• Support get-platform-details for mlperf-inference by @arjunsuresh in #373
• Support system_info.txt in MLPerf inference submission generation by @arjunsuresh in #374
• Cleanups for mlperf inference get-platform-details by @arjunsuresh in #375
• Improvements to get-platform-details by @arjunsuresh in #376
• Improvements for reproducing AMD implementation by @anandhu-eng in #379
• Improvements to amd LLAMA2 70B command generation - Added server scenario by @anandhu-eng in #383
• Build wheels and release them into PYPI by @anandhu-eng in #385
• Do not pass mlperf_conf for inference-src >= 4.1.1 by @arjunsuresh in #404
• Fix version check for mlperf-inference-src by @arjunsuresh in #405
• Added no-compilation-warning variation for loadgen by @arjunsuresh in #406
• Support 8G Nvidia GPUs for MLPerf Inference by @arjunsuresh in #411
• Fix bug on benchmark-program exit check by @arjunsuresh in #412
• Improve the benchmark-program-mlperf run command by @arjunsuresh in #413
• CM4MLOps snapshot with MLPerf inference: 20241024 by @gfursin in #415
• fix batch size duplication issue by @anandhu-eng in #416
• Update cm repo branch - docker by @anandhu-eng in #422
• Fixes for Nvidia MLPerf inference SS and MS by @arjunsuresh in #423
• Improvements for Nvidia MLPerf inference by @arjunsuresh in #428
• added compressed_tools module by @anandhu-eng in #430
• Updated logic for mounting non cache folder by @anandhu-eng in #427
• Fixes for latest MLPerf inference submission checker changes by @arjunsuresh in #431
• Fixes for Latest MLPerf inference changes by @arjunsuresh in #432
• Fixes for latest MLPerf inference changes by @arjunsuresh in #433
• Submission generation fixes by @anandhu-eng in #424
• Support custom path for saving platform details by @anandhu-eng in #418
• Add getting started to cm4mlops docs by @anandhu-eng in #435
• Merge from Mlperf inference by @anandhu-eng in #436
• Fixes for docker detached mode by @arjunsuresh in #438
• Fixes for get-platform-details by @arjunsuresh in #441
• Testing CM Test automation by @arjunsuresh in #442
• Added github action for individual CM tests by @arjunsuresh in #443
• Fix Individual CM script test by @arjunsuresh in #445
• Fix individual CM script test by @arjunsuresh in #446
• Fix gh action for individual CM sript tests by @arjunsuresh in #447
• Initial PR - gh actions for submission generation for non CM based benchmarks by @anandhu-eng in #440
• Fix GH action for individual CM script testing by @arjunsuresh in #449
• Enable docker run for individual CM script tests by @arjunsuresh in #450
• Fixes huggingface downloader by @arjunsuresh in #452
• capture framework version from cm_sut_info.json by @anandhu-eng in #451
• Sync: Mlperf inference by @arjunsuresh in #444
• Support docker_base_image and docker_cm_repo for CM tests by @arjunsuresh in #453
• Improved docker meta for cm test script by @arjunsuresh in #454
• Support test_input_index and test_input_id to customise CM test script by @arjunsuresh in #458
• Improvements to version-detect in get-generic-sys-util by @arjunsuresh in #460
• Use pkg-config deps for get-generic-sys-util by @arjunsuresh in #464
• Fix pstree version detect on macos by @arjunsuresh in #466
• Fix Nvidia mlperf inference retinanet | onnx version by @arjunsuresh in #468
• Support detached mode for nvidia-mlperf-inference-gptj by @arjunsuresh in #469
• Cleanups for get-generic-sys-util by @arjunsuresh in #470
• Fix tmp-run-env.out name by @arjunsuresh in #471
• Add Version RE for g++-11 by @arjunsuresh in #472
• Use berkeley link for imagenet-aux by default by @arjunsuresh in #473
• Code cleanup by @anandhu-eng in #475
• Fixes for Sdxl MLPerf inference by @arjunsuresh in #481
• Added google dns to mlperf-inference docker by @arjunsuresh in #484
• enable submission generation gh action test globally by @anandhu-eng in #487
• Enables docker run in inference submission generation by @anandhu-eng in #486
• Added docker detatched option by @anandhu-eng in #477
• Fixes #261 partially by @Oseltamivir in #426
• Enabled docker run - gh action submission generation by @anandhu-eng in #489
• Fixes for MLPerf inference, intel conda URL by @arjunsuresh in #491
• Dont use '-dt' for Nvidia ml-model-gptj by @arjunsuresh in #492
• Update starting weights filename for SDXL MLPerf inference by @arjunsuresh in #494
• Implements #455: Copy local repo to docker instead of git clone by @Oseltamivir in #467
• Fixes for Nvidia MLPerf inference gptj,sdxl by @arjunsuresh in #495
• Cleanups to MLPerf inference preprocess script by @arjunsuresh in #496
• Support sample-ids for coco2014 accuracy script by @arjunsuresh in #497
• Support download to host - amd llama2 by @anandhu-eng in #480
• Added a retry for git clone failure by @arjunsuresh in #499
• Use custom version for dev branch of inference-src by @arjunsuresh in #500
• Fix path error shown to user by @anandhu-eng in #502
• Fixes for the MLPerf inference nightly test failures by @arjunsuresh in #506
• pip install cm4mlops - handle systems where sudo is absent by @anandhu-eng in #504
• logic update for detect sudo by @anandhu-eng in #508
• Fix Nvidia MLPerf inference gptj model name suffix by @arjunsuresh in #509
• Skip sys-utils-install when no sudo is available by @arjunsuresh in #512
• Fixes against issue: #510 by @anandhu-eng in #514
• Support for privileged mode in docker by @anandhu-eng in #511
• Allow default_version update inside variations by @arjunsuresh in #515
• MLPerf inference changes by @arjunsuresh in #518
• Removed dev branch for SDXL by @arjunsuresh in #519
• update_env_if_env -> update_meta_if_env by @arjunsuresh in #520
• Automatically generate system_meta.json if not found by @anandhu-eng in #523
• Add cuda information to system meta by @anandhu-eng in #526
• Improvements to github actions by @arjunsuresh in #527
• Fix README generation in MLPerf inference submission by @arjunsuresh in #529
• Avoid creation of empty model_mapping.json by @arjunsuresh in #530
• Enhancements to Docker for Multi-User Setups by @arjunsuresh in #539
• Enable update mlperf inference measurements.json file from existing file by @arjunsuresh in #540
• Github action updates by @arjunsuresh in #548
• Enable creating test dataset from main mixtral dataset by @anandhu-eng in #550
• Added RTX4090x1 configs by @arjunsuresh in #551
• change in variation name by @anandhu-eng in #552
• Fix skip model for MLPerf inference llama2 and mixtral by @arjunsuresh in #553
• Support min_query_count and max_query_count for mlperf inference by @arjunsuresh in #554
• Remove TEST05 for MLPerf inference by @arjunsuresh in #555
• Fix CM_ML_MODEL_PATH export for mixtral by @arjunsuresh in #556
• Fix the saving of console logs in benchmark-program - fixes #533 by @arjunsuresh in #558
• Fix 3dunet SS latency in configs by @arjunsuresh in #559
• Added sympy dependency for nvidia mlperf inference gptj by @arjunsuresh in #560
• Added new docker image name for Nvidia MLPerf inference LLM models by @arjunsuresh in #562
• Build TRTLLM for Nvidia MLPerf LLM models by @arjunsuresh in #563
• Update default-config.yaml by @arjunsuresh in #564
• fix get_cudnn and get_tensorrt not detecting multi-digit version numbers correctly by @Submandarine in #565
• add gnn dataset download script by @anandhu-eng in #570
• rename mixtral dataset download script by @anandhu-eng in #569
• enable user to submit a result for both closed and open by @anandhu-eng in #535
• use the default sut folder name supplied if cm-sut-json is not there by @anandhu-eng in #537
• Added dependency graph for mlperf-inference runs by @arjunsuresh in #573
• Add get-rocm-devices for AMD GPUs by @Henryfzh in #544
• Add aarch64-Compatible Base Image for MLPerf Inference & Fix TensorRT Version Matching by @Leonard226 in #478
• additional tag for dataset sampling - mixtral by @anandhu-eng in #576
• Update nvidia mlperf inference sdxl configs by @arjunsuresh in #578
• Improve docker detached mode error capture by @arjunsuresh in #579
• Use absolute paths for docker mounts by @arjunsuresh in #580
• Issue #542 - folder structure rearrange - submission generation by @anandhu-eng in #574
• Fixes for SDXL accuracy run by @arjunsuresh in #582
• Fix a typo in benchmark-program by @arjunsuresh in #583
• _cm.json -> _cm.yaml by @arjunsuresh in #584
• Sync <- Mlperf inference for November release by @arjunsuresh in #585
• Support for platform details in submission generation by @anandhu-eng in #581
• Add remaining gh actions for cm submission generation by @anandhu-eng in #587

New Contributors

• @Oseltamivir made their first contribution in #426
• @Submandarine made their first contribution in #565
• @Henryfzh made their first contribution in #544
• @Leonard226 made their first contribution in #478

Full Changelog: r20241005a...cm4mlperf-v2.3.5