---
library_name: kvpress
license: cc-by-nc-4.0
pipeline_tag: text-generation
tags:
- kv-cache-compression
- long-context
- lora
---

# RestoreKV Checkpoints

Trained restore-token embeddings + LoRA adapters for
**RestoreKV** ([paper](https://arxiv.org/abs/2608.01247) | [project page](https://paper.pnu-cvsp.com/RestoreKV/)), a budget-matched,
single-pass plug-in that recovers full-cache behavior under aggressive
query-agnostic KV cache eviction (built on
[KVzip](https://github.com/snu-mllab/KVzip)).

Each checkpoint stores only the 8 restore-token embeddings and the rank-8 LoRA
adapters (~0.4% of the backbone; ~41 MB). The base model weights are **not**
included — load them from their original repositories.

| File | Backbone | Base scorer |
|---|---|---|
| `llama3.1-8b_restorekv.pt`      | meta-llama/Llama-3.1-8B-Instruct | KVzip  |
| `llama3.1-8b_restorekv_plus.pt` | meta-llama/Llama-3.1-8B-Instruct | KVzip+ |
| `qwen3-8b_restorekv.pt`         | Qwen/Qwen3-8B                    | KVzip  |
| `qwen3-8b_restorekv_plus.pt`    | Qwen/Qwen3-8B                    | KVzip+ |

## Usage (KVPress)

```python
from kvpress import RestoreKVPress
press = RestoreKVPress(
    compression_ratio=0.9375,                       # 16x
    checkpoint_path="llama3.1-8b_restorekv.pt",     # downloaded from this repo
    budget_matched=True,
)
```

Evaluate with the official KVPress protocol (RULER-4096, Llama-3.1-8B):

```bash
python evaluate.py --dataset ruler --data_dir 4096 \
  --model meta-llama/Llama-3.1-8B-Instruct \
  --press_name RestoreKV --compression_ratio 0.875 \
  --restore_checkpoint llama3.1-8b_restorekv.pt
```

For the `*_plus.pt` checkpoints, use the KVzip+ variant
(`--press_name RestoreKV_plus`).

## License

Released under **CC BY-NC 4.0** (research / non-commercial use only). These
checkpoints are distilled using a training mixture that contains
[LongAlpaca](https://huggingface.co/datasets/Yukang/LongAlpaca-12k)-derived
self-study data, so they inherit its most restrictive component license
(CC BY-NC 4.0).