# DataLossError: 无法打开检查点文件 /path/to/checkpoint: 数据损坏: 文件已损坏

- **ID:** `tensorflow/checkpoint-read-failed-corrupt`
- **领域:** tensorflow
- **类别:** data_error
- **错误码:** `CKPT_CORRUPT`
- **验证级别:** ai_generated
- **修复率:** 70%

## 根因

检查点文件由于写入不完整、磁盘故障或 TensorFlow 版本不兼容而损坏。

## 版本兼容性

| 版本 | 状态 | 引入 | 弃用 |
|------|------|------|------|
| tensorflow 2.8.0 | active | — | — |
| tensorflow 2.9.0 | active | — | — |
| tensorflow 2.10.0 | active | — | — |

## 解决方案

1. ```
   Restore from a backup checkpoint or re-train the model. If partial recovery is possible, use `tf.train.load_checkpoint` to inspect and salvage non-corrupted tensors.
   ```

## 无效尝试

- **** — If the source itself is corrupted, re-downloading won't help. (80% 失败率)
- **** — Corruption is usually at the file level, not format level. (85% 失败率)