143 lines
5.4 KiB
Bash
Executable File
143 lines
5.4 KiB
Bash
Executable File
#!/usr/bin/env bash
|
||
set -euo pipefail
|
||
|
||
# ===== 可调参数 =====
|
||
CKPT_ROOT="/home/test/checkpoints/q3-32b-ds4" # 若实际是 .../checkpoint-62/global_step62，请把 CKPT_ROOT 改成 .../checkpoint-62
|
||
TAG="global_step62"
|
||
HOSTS=(tn01 tn02 tn03 tn04 tn05 tn06)
|
||
OUT_DIR="${CKPT_ROOT}/merged-${TAG}"
|
||
MAX_SHARD_SIZE="5GB"
|
||
|
||
# 预检控制：总分片数（通常=总写出 rank 数，如 6 节点×4 GPU=24），最小每机分片数（按你布局调）
|
||
EXPECTED_TOTAL_SHARDS=24
|
||
MIN_SHARDS_PER_HOST=1
|
||
STRICT_PRECHECK=true # true: 若总分片≠期望则直接退出；false: 仅告警但继续
|
||
|
||
SSH_OPTS="-o BatchMode=yes -o StrictHostKeyChecking=accept-new -o ConnectTimeout=8"
|
||
RSYNC_OPTS="-a --info=progress2 --human-readable --partial --inplace"
|
||
# ====================
|
||
|
||
export OUT_DIR # 让后面的 Python 自检拿得到
|
||
|
||
echo "== 预检查 SSH =="
|
||
for h in "${HOSTS[@]}"; do
|
||
ssh ${SSH_OPTS} "$h" "true" >/dev/null || { echo "!! 无法免密 SSH 到 $h"; exit 1; }
|
||
done
|
||
|
||
echo "== 0/4 逐节点分片预检（只统计 ${CKPT_ROOT}/${TAG} 下的 *model_states.pt 文件） =="
|
||
total=0
|
||
declare -A host_cnt
|
||
for h in "${HOSTS[@]}"; do
|
||
# 只认文件，不认目录；限制在 TAG 这一层
|
||
c=$(ssh ${SSH_OPTS} "$h" "find '${CKPT_ROOT}/${TAG}' -maxdepth 1 -type f -name '*model_states.pt' 2>/dev/null | wc -l" || echo 0)
|
||
c=$(echo "$c" | tr -d ' ')
|
||
host_cnt["$h"]=$c
|
||
total=$(( total + c ))
|
||
printf " - %-8s: %s 分片\n" "$h" "$c"
|
||
if (( c < MIN_SHARDS_PER_HOST )); then
|
||
echo "!! 预警：$h 分片仅 $c 个 (< ${MIN_SHARDS_PER_HOST})，该节点可能未写出或路径不同" >&2
|
||
fi
|
||
done
|
||
echo " - 汇总分片数（未同步前）：$total"
|
||
|
||
if [[ -n "${EXPECTED_TOTAL_SHARDS:-}" ]]; then
|
||
if (( total != EXPECTED_TOTAL_SHARDS )); then
|
||
echo "!! 分片总数($total) ≠ 期望(${EXPECTED_TOTAL_SHARDS})，很可能缺片或路径不一致" >&2
|
||
if [[ "${STRICT_PRECHECK}" == "true" ]]; then
|
||
echo "!! STRICT_PRECHECK 开启：中止合并，请先排查缺片节点" >&2
|
||
exit 2
|
||
else
|
||
echo ">> 严格校验关闭：继续执行（可能在合并/加载时失败）" >&2
|
||
fi
|
||
fi
|
||
fi
|
||
|
||
echo "== 1/4 同步各节点的 ${TAG} 目录（带进度）=="
|
||
mkdir -p "${CKPT_ROOT}/${TAG}"
|
||
LOCAL_HOST="$(hostname -s || hostname)"
|
||
for h in "${HOSTS[@]}"; do
|
||
if [[ "$h" == "$LOCAL_HOST" ]]; then
|
||
echo " - 跳过本机 $h"
|
||
continue
|
||
fi
|
||
if ssh ${SSH_OPTS} "$h" "test -d '${CKPT_ROOT}/${TAG}'"; then
|
||
echo " - 从 $h 拉取 ${CKPT_ROOT}/${TAG}/"
|
||
rsync ${RSYNC_OPTS} -e "ssh ${SSH_OPTS}" \
|
||
"${h}:${CKPT_ROOT}/${TAG}/" "${CKPT_ROOT}/${TAG}/" || true
|
||
else
|
||
echo " - $h 无 ${CKPT_ROOT}/${TAG}，跳过"
|
||
fi
|
||
done
|
||
|
||
echo "== 2/4 统计与校验分片文件（本机聚合后） =="
|
||
# 只认文件，不认目录；用一次 find 去重，避免重复计数
|
||
mapfile -t SHARDS < <(find "${CKPT_ROOT}/${TAG}" -maxdepth 1 -type f -name "*model_states.pt" | sort -u)
|
||
CNT=${#SHARDS[@]}
|
||
echo " - 发现 model_states 分片文件数：${CNT}"
|
||
if [[ "${CNT}" -eq 0 ]]; then
|
||
echo "!! 未检测到任何 *model_states.pt；请在各机上 ls 看看 ${CKPT_ROOT}/${TAG} 的实际文件名" >&2
|
||
exit 1
|
||
fi
|
||
# 简单健壮性检查：分片数至少不低于主机数（经验性检查）
|
||
if [[ "${CNT}" -lt "${#HOSTS[@]}" ]]; then
|
||
echo "!! 分片数(${CNT}) < 主机数(${#HOSTS[@]})，可能有节点没同步到分片，继续可能失败" >&2
|
||
fi
|
||
|
||
echo "== 3/4 合并为 safetensors 到：${OUT_DIR} =="
|
||
mkdir -p "${OUT_DIR}"
|
||
|
||
# 先探测 zero_to_fp32.py 是否支持新参数；不支持就走 API
|
||
USE_Z2FP32_SCRIPT=false
|
||
if [[ -f "${CKPT_ROOT}/zero_to_fp32.py" ]]; then
|
||
if python "${CKPT_ROOT}/zero_to_fp32.py" --help 2>&1 | grep -q -- "--safe_serialization"; then
|
||
USE_Z2FP32_SCRIPT=true
|
||
fi
|
||
fi
|
||
|
||
if $USE_Z2FP32_SCRIPT; then
|
||
python "${CKPT_ROOT}/zero_to_fp32.py" \
|
||
"${CKPT_ROOT}" \
|
||
"${OUT_DIR}" \
|
||
--tag "${TAG}" \
|
||
--safe_serialization \
|
||
--max_shard_size "${MAX_SHARD_SIZE}"
|
||
else
|
||
python - <<PY
|
||
from deepspeed.utils.zero_to_fp32 import convert_zero_checkpoint_to_fp32_state_dict
|
||
convert_zero_checkpoint_to_fp32_state_dict(
|
||
checkpoint_dir=r"${CKPT_ROOT}",
|
||
output_dir=r"${OUT_DIR}",
|
||
tag=r"${TAG}",
|
||
safe_serialization=True,
|
||
max_shard_size=r"${MAX_SHARD_SIZE}",
|
||
)
|
||
print("合并完成：", r"${OUT_DIR}")
|
||
PY
|
||
fi
|
||
|
||
echo "== 3.1 拷贝 config/tokenizer 工件（如存在）=="
|
||
for f in config.json generation_config.json tokenizer_config.json tokenizer.json merges.txt vocab.json special_tokens_map.json added_tokens.json; do
|
||
[[ -f "${CKPT_ROOT}/${f}" ]] && cp -n "${CKPT_ROOT}/${f}" "${OUT_DIR}/"
|
||
done
|
||
|
||
echo "== 4/4 自检（索引与 config）=="
|
||
python - <<'PY'
|
||
import os, json, sys
|
||
out_dir = os.environ.get("OUT_DIR")
|
||
idx = os.path.join(out_dir, "model.safetensors.index.json")
|
||
if os.path.exists(idx):
|
||
with open(idx) as f: j = json.load(f)
|
||
print(f"OK: 找到 safetensors 索引：{idx}（参数条目 {len(j.get('weight_map', {}))}）")
|
||
else:
|
||
# 单分片时没有 index.json，属于正常情况
|
||
print("NOTE: 未找到 model.safetensors.index.json（可能是单分片）")
|
||
try:
|
||
from transformers import AutoConfig
|
||
cfg = AutoConfig.from_pretrained(out_dir)
|
||
print("OK: 读取到 config：", cfg.model_type, "hidden:", getattr(cfg,'hidden_size',None), "layers:", getattr(cfg,'num_hidden_layers',None))
|
||
except Exception as e:
|
||
print("WARN: 读取 config 失败（若无 config.json 可忽略）：", e, file=sys.stderr)
|
||
PY
|
||
|
||
echo "== 完成：${OUT_DIR} =="
|