# DataLossError: Unable to open table file /path/to/checkpoint: Data loss: file is corrupted

- **ID:** `tensorflow/checkpoint-read-failed-corrupt`
- **Domain:** tensorflow
- **Category:** data_error
- **Error Code:** `CKPT_CORRUPT`
- **Verification:** ai_generated
- **Fix Rate:** 70%

## Root Cause

The checkpoint file is corrupted due to incomplete write, disk failure, or version incompatibility between TensorFlow versions.

## Version Compatibility

| Version | Status | Introduced | Deprecated |
|---------|--------|------------|------------|
| tensorflow 2.8.0 | active | — | — |
| tensorflow 2.9.0 | active | — | — |
| tensorflow 2.10.0 | active | — | — |

## Workarounds

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.** (70% success)
   ```
   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.
   ```

## Dead Ends

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