script/hf_update.py

from huggingface_hub import HfApi
import os
import zipfile
from tqdm import tqdm

TOKEN = os.environ.get("HF_TOKEN", "")
api = HfApi(token=TOKEN)

base = r"D:\desktop\paper\abs"
repo_id = "zzhowe1207/tet"

# 创建仓库
api.create_repo(repo_id=repo_id, repo_type="dataset", exist_ok=True)

# ========== 1. 分卷压缩数据集 (每卷1GB) ==========
print(">>> 扫描数据集文件...")
data_path = os.path.join(base, "data")

# 收集所有文件
all_files = []
for root, dirs, files in os.walk(data_path):
    for file in files:
        fpath = os.path.join(root, file)
        all_files.append((fpath, os.path.relpath(fpath, base)))

print(f"共 {len(all_files)} 个文件")

# 创建完整zip，带进度
full_zip = os.path.join(base, "data_full.zip")
print("\n>>> 压缩数据集...")
with zipfile.ZipFile(full_zip, 'w', zipfile.ZIP_DEFLATED) as zipf:
    for fpath, arcname in tqdm(all_files, desc="压缩进度", unit="file"):
        zipf.write(fpath, arcname)

full_size = os.path.getsize(full_zip)
print(f"\n总大小: {full_size/1024/1024/1024:.2f} GB")

# 分卷 1GB
volume_size = 1024 * 1024 * 1024  # 1GB
volume_num = (full_size + volume_size - 1) // volume_size
print(f"分成 {volume_num} 卷 (每卷1GB)")

print("\n>>> 生成分卷...")
with open(full_zip, 'rb') as f:
    for i in tqdm(range(volume_num), desc="分卷进度", unit="vol"):
        chunk = f.read(volume_size)
        vol_path = os.path.join(base, f"data.zip.{i+1:03d}")
        with open(vol_path, 'wb') as vol:
            vol.write(chunk)

os.remove(full_zip)

# 上传所有分卷，带进度
print("\n>>> 上传分卷到 Hugging Face...")
volumes = [os.path.join(base, f"data.zip.{i+1:03d}") for i in range(volume_num)]

for i, vol_path in enumerate(tqdm(volumes, desc="上传进度", unit="vol")):
    vol_name = f"data.zip.{i+1:03d}"
    api.upload_file(
        path_or_fileobj=vol_path,
        path_in_repo=vol_name,
        repo_id=repo_id,
        repo_type="dataset"
    )
    os.remove(vol_path)
    print(f"  ✓ {vol_name} 完成")

print("\n数据集分卷上传完成！")

# ========== 2. 不压缩，直接传代码 ==========
print("\n>>> 上传 LLaMA-Factory...")
api.upload_folder(
    folder_path=os.path.join(base, "LLaMA-Factory"),
    path_in_repo="LLaMA-Factory",
    repo_id=repo_id,
    repo_type="dataset"
)
print("  ✓ LLaMA-Factory 完成")

print(">>> 上传 script...")
api.upload_folder(
    folder_path=os.path.join(base, "script"),
    path_in_repo="script",
    repo_id=repo_id,
    repo_type="dataset"
)
print("  ✓ script 完成")

print("\n" + "="*50)
print("搞定！")
print("="*50)
print("\n合并命令: copy /b data.zip.001 + data.zip.002 + ... data.zip")