Skip to content

Latest commit

 

History

History
21 lines (17 loc) · 838 Bytes

File metadata and controls

21 lines (17 loc) · 838 Bytes

NL2Shell Training — Program Constraints

Budget

  • Max compute: <15 compute units (Google Colab A100)
  • Expected: ~3-5 CU for 3 epochs on NL2Bash (~10k examples)

Metrics

  • Primary: eval_loss (lower is better)
  • Qualitative: 7 NL->shell test prompts in prepare.py:EVAL_PROMPTS
  • Success: model produces syntactically valid shell commands for >=5/7 prompts

Target

Rules

  1. Do NOT modify prepare.py — it is immutable
  2. Edit only train.py for hyperparameter tuning or bug fixes
  3. Close Colab session when training completes
  4. If training loss plateaus, reduce learning rate or increase epochs
  5. If OOM, reduce batch_size from 8 to 4 (keep grad_accum=4)