[Bug]: kv_cache_cpu patch reads retired DynamicCache attributes, AttributeError on current transformers

[AmirF194]

Describe the bug

minference_patch_kv_cache_cpu (patch.py) monkeypatches transformers.cache_utils.DynamicCache.update/.get with cpu_cache_update/cpu_cache_get. Both read and write self.key_cache, self.value_cache and self._seen_tokens/self.seen_tokens directly. Current transformers moved DynamicCache's per-layer state onto self.layers and no longer defines those attributes, so the first cache update after kv_cache_cpu=True raises AttributeError.

Same root cause as #207 (fixed for DynamicCacheWithRepeat in #216), but a different call site that fix didn't touch. kv_cache_cpu is a documented MInferenceConfig option used across the example/benchmark scripts, not a corner case.

Steps to reproduce

No vllm or model load needed, this extracts the two patched functions and runs them against a real DynamicCache:

import torch
from transformers.cache_utils import DynamicCache
# cpu_cache_update copied verbatim from minference/patch.py
cache = DynamicCache()
hasattr(cache, "key_cache")  # False
cpu_cache_update(cache, torch.randn(1, 2, 3, 4), torch.randn(1, 2, 3, 4), layer_idx=0)

Expected Behavior

The cache update should succeed (or the patch should adapt to the cache layout transformers actually ships).

Logs

AttributeError: 'DynamicCache' object has no attribute 'seen_tokens'

Same AttributeError from cpu_cache_get.

Additional Information

Confirmed on current main (29ef197) against transformers==5.17.0. Happy to send a PR if useful.