ml-intern-explorers/efficient-optimizer-collab / artifacts /adamw_tuned_cmpatino-0 /run_validation.sh
| # Full-length 5625-step validation of best (lr, wd) from sweeps. | |
| set -euo pipefail | |
| cd "$(dirname "$0")/../../modded-nanogpt" | |
| SCRIPT="../artifacts/adamw_baseline_v2_cmpatino-0/train_gpt_adamw_v2_cmpatino-0.py" | |
| OUT="../artifacts/adamw_tuned_cmpatino-0/log_lr0.0010_wd0.05_5625_cmpatino-0.txt" | |
| export TRAIN_STEPS=5625 | |
| export WARMUP_STEPS=250 | |
| export COOLDOWN_FRAC=0.7 | |
| export BLOCK_LR=0.0010 | |
| export BLOCK_WD=0.05 | |
| export BLOCK_BETA1=0.9 | |
| export BLOCK_BETA2=0.95 | |
| export RUN_TAG="adamw_tuned_lr0.0010_wd0.05_cmpatino-0" | |
| ../.venv/bin/torchrun --standalone --nproc_per_node=2 "$SCRIPT" 1 2>&1 | tee "$OUT" | |
Xet Storage Details
- Size:
- 618 Bytes
- Xet hash:
- 8cdc6a58ab282383dd8ac7d1ff90213efac8554f52bccf87cadf65bdcd80278e
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.