Fix hf_olmo GPU tests and transformers>=4.45 compatibility

### 🐛 Describe the bug

I was setting up the repo before noticing recent deprecation notice. Found couple of minor issues causing some of the tests to fail. These still might be relevant for anyone using OLMo 2:

1. GPU tests fail due to device placement bugs in tests/hf_olmo/
2. transformers>=4.45 breaks generation due to DynamicCache API change [source](https://huggingface.co/docs/transformers/en/cache_explanation#legacy-cache-format)

I have fixes ready - tests pass but not validated beyond that.

If useful happy to create PR. Otherwise will focus on olmo-core instead.

### Versions

Python 3.10.19
accelerate==1.12.0
-e file:///home/user/Olmo/OLMo-claude
ai2-olmo-core==2.4.0
aiohappyeyeballs==2.6.1
aiohttp==3.13.2
aiosignal==1.4.0
annotated-types==0.7.0
antlr4-python3-runtime==4.9.3
anyio==4.11.0
async-timeout==5.0.1
attrs==25.4.0
backports-tarfile==1.2.0
beaker-gantry==3.3.0
beaker-py==2.5.2
bettermap==1.3.1
black==23.12.1
boltons==25.0.0
boto3==1.41.3
botocore==1.41.3
build==1.3.0
cached-path==1.8.0
cachetools==6.2.2
certifi==2025.11.12
cffi==2.0.0
charset-normalizer==3.4.4
click==8.3.1
click-help-colors==0.9.4
click-option-group==0.5.9
cryptography==46.0.3
dataclass-extensions==0.2.4
datasets==4.4.1
dill==0.4.0
docutils==0.22.3
einops==0.8.1
exceptiongroup==1.3.1
face==24.0.0
filelock==3.20.0
flash-attn @ https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.3/flash_attn-2.8.3+cu12torch2.8cxx11abiTRUE-cp310-cp310-linux_x86_64.whl
frozenlist==1.8.0
fsspec==2025.10.0
ftfy==6.3.1
gitdb==4.0.12
gitpython==3.1.45
glom==24.11.0
google-api-core==2.28.1
google-auth==2.43.0
google-cloud-core==2.5.0
google-cloud-storage==2.19.0
google-crc32c==1.7.1
google-resumable-media==2.8.0
googleapis-common-protos==1.72.0
grpcio==1.76.0
h11==0.16.0
hf-xet==1.2.0
httpcore==1.0.9
httpx==0.28.1
huggingface-hub==0.36.0
id==1.5.0
idna==3.11
importlib-metadata==8.7.0
importlib-resources==6.5.2
iniconfig==2.3.0
isort==5.12.0
jaraco-classes==3.4.0
jaraco-context==6.0.1
jaraco-functools==4.3.0
jeepney==0.9.0
jinja2==3.1.6
jmespath==1.0.1
joblib==1.5.2
keyring==25.7.0
lightning-utilities==0.15.2
markdown-it-py==4.0.0
markupsafe==3.0.3
mdurl==0.1.2
more-itertools==10.8.0
mpmath==1.3.0
msgspec==0.20.0
multidict==6.7.0
multiprocess==0.70.18
mypy==1.3.0
mypy-extensions==1.1.0
necessary==0.4.3
networkx==3.4.2
nh3==0.3.2
numpy==2.2.6
nvidia-cublas-cu12==12.8.4.1
nvidia-cuda-cupti-cu12==12.8.90
nvidia-cuda-nvrtc-cu12==12.8.93
nvidia-cuda-runtime-cu12==12.8.90
nvidia-cudnn-cu12==9.10.2.21
nvidia-cufft-cu12==11.3.3.83
nvidia-cufile-cu12==1.13.1.3
nvidia-curand-cu12==10.3.9.90
nvidia-cusolver-cu12==11.7.3.90
nvidia-cusparse-cu12==12.5.8.93
nvidia-cusparselt-cu12==0.7.1
nvidia-nccl-cu12==2.27.3
nvidia-nvjitlink-cu12==12.8.93
nvidia-nvshmem-cu12==3.3.20
nvidia-nvtx-cu12==12.8.90
omegaconf==2.3.0
packaging==25.0
pandas==2.3.3
pathspec==0.12.1
petname==2.6
platformdirs==4.5.0
pluggy==1.6.0
propcache==0.4.1
proto-plus==1.26.1
protobuf==6.33.1
psutil==7.1.3
pyarrow==22.0.0
pyasn1==0.6.1
pyasn1-modules==0.4.2
pycparser==2.23
pydantic==2.12.4
pydantic-core==2.41.5
pygments==2.19.2
pyproject-hooks==1.2.0
pytest==9.0.1
pytest-sphinx==0.6.3
python-dateutil==2.9.0.post0
pytz==2025.2
pyyaml==6.0.3
readme-renderer==44.0
regex==2025.11.3
requests==2.32.5
requests-toolbelt==1.0.0
requirements-parser==0.13.0
rfc3986==2.0.0
rich==13.9.4
rsa==4.9.1
ruff==0.14.6
s3transfer==0.15.0
safetensors==0.7.0
scikit-learn==1.7.2
scipy==1.15.3
secretstorage==3.5.0
sentry-sdk==2.46.0
setuptools==80.9.0
six==1.17.0
smart-open==7.5.0
smashed==0.21.5
smmap==5.0.2
sniffio==1.3.1
sympy==1.14.0
threadpoolctl==3.6.0
tokenizers==0.22.1
tomli==2.3.0
torch==2.8.0
torchmetrics==1.8.2
tqdm==4.67.1
transformers==4.57.2
triton==3.4.0
trouting==0.3.3
twine==6.2.0
typing-extensions==4.15.0
typing-inspection==0.4.2
tzdata==2025.2
urllib3==2.5.0
wandb==0.23.0
wcwidth==0.2.14
wheel==0.45.1
wrapt==2.0.1
xxhash==3.6.0
yarl==1.22.0
zipp==3.23.0