Add RestoreKV checkpoints (Llama-3.1-8B, Qwen3-8B; base + plus)

README.md

@@ -0,0 +1,50 @@
+---
+license: mit
+tags:
+  - kv-cache-compression
+  - long-context
+  - lora
+library_name: kvpress
+---
+
+# RestoreKV Checkpoints
+
+Trained restore-token embeddings + LoRA adapters for
+**RestoreKV** ([paper](https://arxiv.org/abs/2608.01247)), 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`).

llama3.1-8b_restorekv.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9511ff553f637eae96a8e6284e6a1db11af12a977c7146d242f08c75b8081705
+size 42158929

llama3.1-8b_restorekv_plus.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36134fea178e3a6d7e5d99ae30fbb69ef28491fac7ee44f822654e1ed190d90f
+size 42158929

qwen3-8b_restorekv.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89a1ec90dea5f826aa831bff71d63a8ea43dc6cb55f8030bfe80302f468379f4
+size 43881281

qwen3-8b_restorekv_plus.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb634cff1aab0f0dc44ef7eae24833565fdb792b4e69cd8bba7fdbb69dda2c0f
+size 43881281