UPLOAD.sh

#!/bin/bash
# HF Upload Script — run on VAE devspace after HF_TOKEN is set
# Usage:
#   export HF_TOKEN=hf_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
#   bash HF_UPLOAD_PLAN.sh
set -euo pipefail

: "${HF_TOKEN:?Set HF_TOKEN to an anonymous-account write token}"

STAGE=~/bmvc_hf_release
DATA_REPO_NAMES=(
  "anonymous-bmvc-2026-3d-contamination"
  "bmvc2026-3d-axes"
  "anon-bmvc26-contam-axes"
)

echo "Logging in..."
hf auth login --token "$HF_TOKEN" --add-to-git-credential 2>/dev/null || \
  huggingface-cli login --token "$HF_TOKEN" --add-to-git-credential

WHO=$(hf auth whoami 2>/dev/null || huggingface-cli whoami | head -1)
echo "Logged in as: $WHO"

# Try each name until one is free for both repos
PICKED=""
for NAME in "${DATA_REPO_NAMES[@]}"; do
  DATA_REPO="${WHO}/${NAME}-data"
  CKPT_REPO="${WHO}/${NAME}-checkpoints"
  # Try to create — if 409 (exists), skip to next
  if hf repo create "$DATA_REPO" --repo-type dataset --private false 2>&1 | grep -qiE "already exists|409"; then
    echo "  $DATA_REPO already exists, trying next..."
    continue
  fi
  if hf repo create "$CKPT_REPO" --repo-type model --private false 2>&1 | grep -qiE "already exists|409"; then
    echo "  $CKPT_REPO already exists, trying next..."
    continue
  fi
  PICKED="$NAME"
  break
done

if [ -z "$PICKED" ]; then
  echo "All candidate names were taken. Edit the DATA_REPO_NAMES list and retry."
  exit 1
fi

DATA_REPO="${WHO}/${PICKED}-data"
CKPT_REPO="${WHO}/${PICKED}-checkpoints"
echo "Using: $DATA_REPO (dataset) + $CKPT_REPO (model)"

# Upload data repo (everything except checkpoints/)
echo "=== Uploading data repo ==="
cd $STAGE
# Upload README at root + all non-checkpoint dirs
hf upload "$DATA_REPO" README.md README.md --repo-type dataset
hf upload "$DATA_REPO" masks/ masks/ --repo-type dataset
hf upload "$DATA_REPO" configs/ configs/ --repo-type dataset
hf upload "$DATA_REPO" metrics/ metrics/ --repo-type dataset
hf upload "$DATA_REPO" code/ code/ --repo-type dataset
hf upload "$DATA_REPO" predictions/ predictions/ --repo-type dataset
hf upload "$DATA_REPO" splits/ splits/ --repo-type dataset
hf upload "$DATA_REPO" logs/ logs/ --repo-type dataset
# Also upload claim map + reproduce as top-level docs
hf upload "$DATA_REPO" CLAIM_TO_FILE_MAP.md CLAIM_TO_FILE_MAP.md --repo-type dataset
hf upload "$DATA_REPO" REPRODUCE.md REPRODUCE.md --repo-type dataset
hf upload "$DATA_REPO" CITATIONS.md CITATIONS.md --repo-type dataset
hf upload "$DATA_REPO" LICENSE LICENSE --repo-type dataset

# Upload checkpoint repo (model type for HF inference compatibility)
echo "=== Uploading checkpoint repo ==="
hf upload "$CKPT_REPO" README.md README.md --repo-type model
hf upload "$CKPT_REPO" checkpoints/ . --repo-type model

echo
echo "Done. URLs:"
echo "  Data:        https://huggingface.co/datasets/${DATA_REPO}"
echo "  Checkpoints: https://huggingface.co/${CKPT_REPO}"