Open davidjurado opened 3 years ago
MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅
Hello @davidjurado! I tried to follow the task execution steps, but the last step failed with the following error:
$ mlcube run --task train
Usage: mlcube.py train [OPTIONS]
Try 'mlcube.py train --help' for help.
Error: Missing option '--output_dir'.
2023-05-19 09:35:17 [...]
Your description sais:
# Run benchmark. Default paths = ./workspace/data
# Parameters to override: data_dir=DATA_DIR, output_dir=OUTPUT_DIR, parameters_file=PATH_TO_TRAINING_PARAMS
mlcube run --task train
How to override the output_dir
?
@davidjurado can you answer @mwawrzos 's question. We can merge this accordingly.
Used PR #465 as reference.
Current implementation
We'll be updating this section as we merge MLCube PRs and make new MLCube releases.
Project setup
Dataset
The Librispeech dataset will be downloaded, extracted, and processed. Sizes of the dataset in each step:
Tasks execution
We are targeting pull-type installation, so MLCube images should be available on docker hub. If not, try this: