d2bfeaa983
Adds a tiny alpine pod (check-cache) that runs first on the platform
pool (no autoscaler delay, ~3 sec end-to-end) and probes the
training-data PVC for /data/bin/$SHA/alpha_train. Outputs:
- sha: short SHA used for binary cache keying
- cache: "hit" or "miss"
ensure-binary now has `when: cache == miss` — when the binary is
already cached for the current SHA, the entire ~4.8GB ci-builder
image pull + sccache compile cycle is skipped. Re-runs on the same
SHA now go straight from submission to training in ~30 seconds
instead of ~3 minutes.
train depends on check-cache + ensure-binary; sources the SHA from
check-cache's output (works whether ensure-binary ran or was
skipped — Argo treats `when:` skip as a satisfied dependency).
Submission script (scripts/argo-alpha-perception.sh) now pre-resolves
commit-sha=HEAD to an actual git SHA via `git rev-parse origin/<branch>`
before submission. This lets the alpine check-cache pod work without
installing git in the container.
Also removed the now-stale `ci-training-h100x2|ci-training-h100-sxm`
case branch from the SM-arch detection — those pools no longer exist
post-pool-cleanup commit a252119fd.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
127 lines
4.1 KiB
Bash
Executable File
127 lines
4.1 KiB
Bash
Executable File
#!/usr/bin/env bash
|
|
# Submit the alpha-perception workflow.
|
|
#
|
|
# Trains the stacked Mamba2 -> CfC -> heads perception model on MBP-10
|
|
# from the training-data PVC. Emits alpha_train_summary.json with
|
|
# per-horizon validation AUC to the feature-cache PVC.
|
|
#
|
|
# Defaults match the validated synthetic-overfit smoke config; cluster
|
|
# runs can override for sweep work.
|
|
#
|
|
# Usage:
|
|
# ./scripts/argo-alpha-perception.sh # current HEAD on ml-alpha-phase-a
|
|
# ./scripts/argo-alpha-perception.sh --branch main
|
|
# ./scripts/argo-alpha-perception.sh --epochs 1 --n-train-seqs 1000 # quick smoke
|
|
# ./scripts/argo-alpha-perception.sh --watch
|
|
|
|
set -euo pipefail
|
|
|
|
SHA=HEAD
|
|
BRANCH=$(git symbolic-ref --short HEAD 2>/dev/null || echo "ml-alpha-phase-a")
|
|
GPU_POOL=ci-training-l40s
|
|
EPOCHS=5
|
|
SEQ_LEN=32
|
|
MAMBA2_STATE_DIM=16
|
|
LR_CFC=3e-3
|
|
LR_MAMBA2=1e-3
|
|
N_TRAIN_SEQS=8000
|
|
N_VAL_SEQS=1000
|
|
SEED=0x4242
|
|
WATCH=false
|
|
|
|
usage() {
|
|
cat <<EOF
|
|
Usage: $0 [OPTIONS]
|
|
--sha <commit> Git SHA (default: HEAD on the current branch)
|
|
--branch <branch> Git branch (default: $BRANCH)
|
|
--gpu-pool <pool> GPU pool (default: $GPU_POOL)
|
|
--epochs <n> Training epochs (default: $EPOCHS)
|
|
--seq-len <n> Snapshots per sequence window (default: $SEQ_LEN)
|
|
--mamba2-state-dim <n> Mamba2 SSM state dim (default: $MAMBA2_STATE_DIM)
|
|
--lr-cfc <f> CfC learning rate (default: $LR_CFC)
|
|
--lr-mamba2 <f> Mamba2 learning rate (default: $LR_MAMBA2)
|
|
--n-train-seqs <n> Train sequences per epoch (default: $N_TRAIN_SEQS)
|
|
--n-val-seqs <n> Val sequences per epoch (default: $N_VAL_SEQS)
|
|
--seed <n> Random seed (default: $SEED)
|
|
--watch Follow logs via argo watch
|
|
EOF
|
|
}
|
|
|
|
while [[ $# -gt 0 ]]; do
|
|
case "$1" in
|
|
--sha) SHA="$2"; shift 2 ;;
|
|
--branch) BRANCH="$2"; shift 2 ;;
|
|
--gpu-pool) GPU_POOL="$2"; shift 2 ;;
|
|
--epochs) EPOCHS="$2"; shift 2 ;;
|
|
--seq-len) SEQ_LEN="$2"; shift 2 ;;
|
|
--mamba2-state-dim) MAMBA2_STATE_DIM="$2"; shift 2 ;;
|
|
--lr-cfc) LR_CFC="$2"; shift 2 ;;
|
|
--lr-mamba2) LR_MAMBA2="$2"; shift 2 ;;
|
|
--n-train-seqs) N_TRAIN_SEQS="$2"; shift 2 ;;
|
|
--n-val-seqs) N_VAL_SEQS="$2"; shift 2 ;;
|
|
--seed) SEED="$2"; shift 2 ;;
|
|
--watch) WATCH=true; shift ;;
|
|
-h|--help) usage; exit 0 ;;
|
|
*) echo "Unknown option: $1"; usage; exit 1 ;;
|
|
esac
|
|
done
|
|
|
|
case "$GPU_POOL" in
|
|
ci-training-l40s)
|
|
SM=89 ;;
|
|
ci-training-h100)
|
|
SM=90 ;;
|
|
*)
|
|
echo "Unknown gpu-pool: $GPU_POOL"
|
|
exit 1 ;;
|
|
esac
|
|
|
|
# Resolve commit-sha=HEAD to an actual SHA locally before submission.
|
|
# The in-cluster check-cache pod uses this SHA to look up the binary
|
|
# cache without needing git inside the alpine pod.
|
|
if [ "$SHA" = "HEAD" ]; then
|
|
echo "Resolving HEAD for branch $BRANCH..."
|
|
if ! git rev-parse --git-dir >/dev/null 2>&1; then
|
|
echo "ERROR: not in a git repo; cannot resolve HEAD"
|
|
exit 1
|
|
fi
|
|
git fetch --quiet origin "$BRANCH" 2>/dev/null || true
|
|
SHA=$(git rev-parse "origin/$BRANCH" 2>/dev/null \
|
|
|| git rev-parse "$BRANCH" 2>/dev/null \
|
|
|| git rev-parse HEAD)
|
|
echo " resolved: $SHA"
|
|
fi
|
|
|
|
echo "Submitting alpha-perception workflow..."
|
|
echo " branch: $BRANCH"
|
|
echo " sha: $SHA"
|
|
echo " gpu-pool: $GPU_POOL (sm_$SM)"
|
|
echo " epochs: $EPOCHS"
|
|
echo " seq-len: $SEQ_LEN"
|
|
echo " mamba2-state-dim: $MAMBA2_STATE_DIM"
|
|
echo " lr-cfc: $LR_CFC"
|
|
echo " lr-mamba2: $LR_MAMBA2"
|
|
echo " n-train-seqs: $N_TRAIN_SEQS"
|
|
echo " n-val-seqs: $N_VAL_SEQS"
|
|
echo " seed: $SEED"
|
|
|
|
WATCH_FLAG=""
|
|
if [[ "$WATCH" == "true" ]]; then
|
|
WATCH_FLAG="--watch"
|
|
fi
|
|
|
|
argo submit -n foxhunt --from=wftmpl/alpha-perception \
|
|
-p commit-sha="$SHA" \
|
|
-p git-branch="$BRANCH" \
|
|
-p cuda-compute-cap="$SM" \
|
|
-p gpu-pool="$GPU_POOL" \
|
|
-p epochs="$EPOCHS" \
|
|
-p seq-len="$SEQ_LEN" \
|
|
-p mamba2-state-dim="$MAMBA2_STATE_DIM" \
|
|
-p lr-cfc="$LR_CFC" \
|
|
-p lr-mamba2="$LR_MAMBA2" \
|
|
-p n-train-seqs="$N_TRAIN_SEQS" \
|
|
-p n-val-seqs="$N_VAL_SEQS" \
|
|
-p seed="$SEED" \
|
|
$WATCH_FLAG
|