Support chunked prefill when radix cache is disabled (#811)

python/sglang/srt/constrained/base_tool_cache.py

@@ -13,12 +13,12 @@ See the License for the specific language governing permissions and
 limitations under the License.
 """
 
-"""Base cache class."""
+"""Base tool cache for constrained decoding tools."""
 
 import time
 
 
-class BaseCache:
+class BaseToolCache:
     def __init__(self, enable=True):
         self.enable = enable
         self.reset()

python/sglang/srt/constrained/fsm_cache.py

@@ -16,10 +16,10 @@ limitations under the License.
 """Cache for the compressed finite state machine."""
 
 from sglang.srt.constrained import RegexGuide, TransformerTokenizer
-from sglang.srt.constrained.base_cache import BaseCache
+from sglang.srt.constrained.base_tool_cache import BaseToolCache
 
 
-class FSMCache(BaseCache):
+class FSMCache(BaseToolCache):
     def __init__(self, tokenizer_path, tokenizer_args_dict, enable=True):
         super().__init__(enable=enable)
 

python/sglang/srt/constrained/jump_forward.py

@@ -30,7 +30,7 @@ from sglang.srt.constrained import (
     make_byte_level_fsm,
     make_deterministic_fsm,
 )
-from sglang.srt.constrained.base_cache import BaseCache
+from sglang.srt.constrained.base_tool_cache import BaseToolCache
 
 IP_REGEX = r"((25[0-5]|2[0-4]\d|[01]?\d\d?)\.){3}(25[0-5]|2[0-4]\d|[01]?\d\d?)"
 
@@ -151,7 +151,7 @@ class JumpForwardMap:
         )
 
 
-class JumpForwardCache(BaseCache):
+class JumpForwardCache(BaseToolCache):
     def __init__(self):
         super().__init__()