sgl-project/sglang · critical · RuntimeError

LMCache is not installed. Please install it by running `pip

Error message

LMCache is not installed. Please install it by running `pip install lmcache`

What it means

Module-level import of lmcache.integration.sglang (connectors, LoadMetadata/StoreMetadata, lmcache_get_config) failed with ImportError, which is re-raised as RuntimeError telling the user to pip install lmcache. Since the import is at module scope, merely importing lmc_radix_cache triggers it.

Source

Thrown at python/sglang/srt/mem_cache/storage/lmcache/lmc_radix_cache.py:31

    EvictResult,
    InitLoadBackParams,
    MatchPrefixParams,
    MatchResult,
)
from sglang.srt.mem_cache.radix_cache import RadixCache, RadixKey, TreeNode
from sglang.srt.runtime_context import get_memory, get_spec
from sglang.srt.utils import create_device_stream, device_stream_context

try:
    from lmcache.integration.sglang.multi_process_adapter import LMCacheMPConnector
    from lmcache.integration.sglang.sglang_adapter import (
        LMCacheLayerwiseConnector,
        LoadMetadata,
        StoreMetadata,
    )
    from lmcache.integration.sglang.utils import lmcache_get_config
except ImportError as e:
    raise RuntimeError(
        "LMCache is not installed. Please install it by running `pip install lmcache`"
    ) from e


if TYPE_CHECKING:
    from sglang.srt.configs.model_config import ModelConfig
    from sglang.srt.managers.schedule_batch import Req
    from sglang.srt.mem_cache.cache_init_params import CacheInitParams

logger = logging.getLogger(__name__)


@dataclass
class _LMCacheLoadBackMarker:
    """Carries the data ``init_load_back`` needs from the
    ``match_prefix`` call in MP mode.
    """

View on GitHub (pinned to 0132848349)

Solutions

1. pip install lmcache in the same interpreter that runs SGLang (python -m pip install lmcache)
2. Verify with: python -c "from lmcache.integration.sglang.utils import lmcache_get_config"
3. If that fails with a deeper ImportError, reinstall lmcache matching your Python/torch versions

try:
    from lmcache.integration.sglang.utils import lmcache_get_config  # noqa: F401
    LMCACHE_OK = True
except ImportError:
    LMCACHE_OK = False

assert LMCACHE_OK, 'pip install lmcache in this interpreter'

def lmcache_available() -> bool:
    try:
        import lmcache.integration.sglang  # noqa: F401
        return True
    except ImportError:
        return False

try:
    from sglang.srt.mem_cache.storage.lmcache.lmc_radix_cache import LMCacheRadixCache
except RuntimeError as e:
    if 'LMCache is not installed' in str(e):
        raise SystemExit('install lmcache before enabling --enable-lmcache')
    raise

When it happens

Trigger: Importing LMCacheRadixCache (or any module importing lmc_radix_cache) in a Python environment where lmcache is not installed or its sglang integration extras are missing.

Common situations: Enabling LMCache without installing the package; lmcache installed for a different Python interpreter than the one running the scheduler (common in venv/conda mixups); broken lmcache install where the top package imports but integration submodule does not.

Related errors

• hf3fs_fuse.io is not available. Please install the hf3fs_fus
• Sparse Video Gen 2 attention backend requires svg package to
• Humming quantization requires `humming-kernels`. Please inst
• Please install mooncake by following the instructions at htt
• The fastokens package is required when --tokenizer-backend=f

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/ae61433b4c79b2e5. Report an issue: GitHub.