Update version to v0.1.13 (#280)

python/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sglang"
-version = "0.1.12"
+version = "0.1.13"
 description = "A structured generation langauge for LLMs." 
 readme = "README.md"
 requires-python = ">=3.8"

python/sglang/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.1.12"
+__version__ = "0.1.13"
 
 from sglang.api import *
 from sglang.global_config import global_config

python/sglang/srt/layers/radix_attention.py

@@ -17,7 +17,7 @@ class RadixAttention(nn.Module):
 
         from sglang.srt.managers.router.model_runner import global_server_args_dict
 
-        if global_server_args_dict["enable_flashinfer"]:
+        if global_server_args_dict.get("enable_flashinfer", False):
             self.prefill_forward = self.prefill_forward_flashinfer
             self.extend_forward = self.prefill_forward_flashinfer
             self.decode_forward = self.decode_forward_flashinfer

python/sglang/srt/layers/token_attention.py

@@ -7,7 +7,7 @@ import triton.language as tl
 from sglang.srt.managers.router.model_runner import global_server_args_dict
 from sglang.srt.utils import wrap_kernel_launcher
 
-if global_server_args_dict["attention_reduce_in_fp32"]:
+if global_server_args_dict.get("attention_reduce_in_fp32", False):
     REDUCE_TRITON_TYPE = tl.float32
     REDUCE_TORCH_TYPE = torch.float32
 else:

python/sglang/srt/managers/router/model_runner.py

@@ -222,7 +222,7 @@ class InputMetadata:
         if forward_mode == ForwardMode.EXTEND:
             ret.init_extend_args()
 
-        if global_server_args_dict["enable_flashinfer"]:
+        if global_server_args_dict.get("enable_flashinfer", False):
             ret.init_flashinfer_args(tp_size)
 
         return ret