Skip to content

MODAL_HOW_TO_RUN.md

Latest commit

 

History

History
56 lines (42 loc) · 2.4 KB

MODAL_HOW_TO_RUN.md

File metadata and controls

56 lines (42 loc) · 2.4 KB

Running TemStaPro on Modal

Conda Environment Setup

How to manage conda: https://docs.conda.io/projects/conda/en/stable/user-guide/tasks/manage-conda.html

How to manage environments: https://docs.conda.io/projects/conda/en/stable/user-guide/tasks/manage-environments.html

  • Install Miniconda or Anaconda if needed.
  • Create an isolated environment (Python 3.10+):
    • conda create -n hemeai python=3.10
    • conda activate hemeai
  • Inside the environment, install any project extras you require (e.g. pip install -e .).

Install Modal & Sign In

  • With the environment active: pip install modal
  • Authenticate once per machine: modal token new and complete the browser flow to create the user session.
  • Run commands from the repo root so modal_temstapro.py sits alongside the temstapro/ package.
  • Expect the first run to spend a few minutes provisioning the Modal container image.

Examples

  • Single sequence with default thresholds: modal run modal_temstapro.py --fasta-path data/example.fasta --output-path outputs/example.tsv
  • Batch of short sequences plus plots: modal run modal_temstapro.py --fasta-path tests/data/multiple_short_sequences.fasta --output-path outputs/multi.tsv --per-res-output outputs/per_res.csv --plot-dir outputs/plots

Command Template

modal run modal_temstapro.py \
  --fasta-path path/to/input.fasta \
  --output-path outputs/mean_predictions.tsv \
  [--per-res-output outputs/per_residue.tsv] \
  [--per-segment-output outputs/per_segment.tsv] \
  [--plot-dir outputs/plots] \
  [--more-thresholds] [--curve-smoothening] \
  [--portion-size 1000] [--segment-size 41] [--window-size-predictions 81]
  • --output-path / --per-res-output / --per-segment-output save Modal results locally; directories are created automatically.
  • --plot-dir downloads any per-residue plots generated remotely.
  • --more-thresholds and --curve-smoothening forward the matching TemStaPro CLI flags.
  • Chunking parameters (--portion-size, --segment-size, --window-size-predictions) are optional overrides for edge cases.

GPU & Timeout Tweaks

  • Set GPU to request hardware (default is CPU):
    • GPU=A10G modal run …
    • Supported values: A10G, A100, H100, L4.
  • Extend Modal's execution time limit via TIMEOUT (minutes): TIMEOUT=240 modal run ….

Outputs are written once the remote job finishes; stdout/stderr from TemStaPro stream back to your terminal.