Conversation
|
Fixing the formatting to match commit checks has caused 144 changed file instead of 3. I can either revert that last commit that did this and do PR without it, or we can stick to current formatting in CI. |
There was a problem hiding this comment.
Pull Request Overview
Refactor of the conversion workflow to be modular, driven via CLI, with environment and job orchestration added.
- Add Micromamba setup script and Kubernetes job spec for reproducible environments and training runs.
- Introduce a timing log (
time.txt) to record processing duration. - Clean up whitespace in
loss.py.
Reviewed Changes
Copilot reviewed 89 out of 144 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| micromamba_setup.sh | Script to create and activate the Conda-like environment |
| models/quantized-dense-embedding/time.txt | Log file capturing working time |
| loss.py | Removed stray blank line |
| l1metml-job2.yml | Kubernetes Job manifest for training |
Comments suppressed due to low confidence (2)
models/quantized-dense-embedding/time.txt:1
- [nitpick] Add a newline or clear delimiter between the seconds and minutes values for readability. For example:
Working Time (s): 19636.02
Working Time (m): 327.27
Working Time (s) : 19636.019178152084Working Time (m) : 327.26698630253475
l1metml-job2.yml:16
- [nitpick] Consider using YAML's block scalar (
|) for this long command to improve readability and maintain consistent indentation, for example:
command: |
git clone ... && \
cd L1METML && \
python train.py ... python train.py --workflowType dataGenerator --input /home/users/dprimosc/data/l1_trigger_ntuples/TTbar --mode 1 --epochs 500 --maxNPF 100 --batch-size 256 --units 12 36 --output models/quantized-dense-embedding/ --quantized 8 2 --model dense_embedding --compute-edge-feat 0 --model-output models/saved_keras_models/quantized_dense_model_100pf_500epochs_1_test --normFac 1"
| @@ -0,0 +1,2 @@ | |||
| micromamba create --file environment.yml --name l1metml | |||
There was a problem hiding this comment.
Consider adding the "-y" flag to the micromamba create command to avoid interactive prompts, e.g., micromamba create -y --file environment.yml --name l1metml.
| micromamba create --file environment.yml --name l1metml | |
| micromamba create -y --file environment.yml --name l1metml |
Refactored convert_full_model.py to be modular
Now runs via CLI with arguments
Tested and bugs fixed
Added hls_conversion_config.yaml (not implemented yet)