This Python script originated from automatically converting the Jupyter Notebook that dividiti used for generating data for the interactive MLPerf Inference v0.5 dashboard:
$ jupyter nbconvert --to script results.ipynb
It can be compared with the original script as follows:
$ diff ../../jnotebook/mlperf-inference-v0.5/results.py ./run.py
$ ck run ck-mlperf:program:dump-submissions-to-dashboard \
--env.CK_MLPERF_SUBMISSION_ROOT=... \
--env.CK_MLPERF_DASHBOARD_FILE=... \
--env.CK_MLPERF_DASHBOARD_DIR=...
where:
- CK_MLPERF_SUBMISSION_ROOT
is the path to an input repository with submissions (e.g. cloned from https://github.com/mlperf/inference_results_v0.5);
- CK_MLPERF_DASHBOARD_FILE
is the name of an output dashboard file (mlperf-inference-v0.5-results.zip
, by default).
- CK_MLPERF_DASHBOARD_DIR
is the name of an output dashboard directory. If empty (by default), dump into in the dashboard module directory (ck find ck-mlperf:module:mlperf.inference
).
Dump mlperf-inference-v0.5-results.zip
from $HOME/projects/mlperf/inference_results_v0.5_dividiti
into the dashboard directory:
$ ck run ck-mlperf:program:dump-submissions-to-dashboard
Dump mlperf-inference-unofficial-results.zip
from $HOME/projects/mlperf/inference_results_v0.5_plus
into the dashboard directory:
$ ck run ck-mlperf:program:dump-submissions-to-dashboard \
--env.CK_MLPERF_SUBMISSION_ROOT=$HOME/projects/mlperf/inference_results_v0.5_plus \
--env.CK_MLPERF_DASHBOARD_FILE=mlperf-inference-unofficial-results.zip
Dump mlperf-inference-unofficial-results.zip
from $HOME/projects/mlperf/inference_results_v0.5_plus
into the current directory:
$ ck run ck-mlperf:program:dump-submissions-to-dashboard \
--env.CK_MLPERF_SUBMISSION_ROOT=$HOME/projects/mlperf/inference_results_v0.5_plus \
--env.CK_MLPERF_DASHBOARD_FILE=mlperf-inference-unofficial-results.zip \
--env.CK_MLPERF_DASHBOARD_DIR=`pwd`