[misc] remove is_cuda_available (#5319)

python/sglang/srt/speculative/build_eagle_tree.py

@@ -4,9 +4,9 @@ from typing import List
 
 import torch
 
-from sglang.srt.utils import is_cuda_available, is_hip
+from sglang.srt.utils import is_cuda, is_hip
 
-if is_cuda_available() or is_hip():
+if is_cuda() or is_hip():
     from sgl_kernel import (
         build_tree_kernel_efficient as sgl_build_tree_kernel_efficient,
     )

python/sglang/srt/speculative/eagle_utils.py

@@ -19,9 +19,9 @@ from sglang.srt.managers.schedule_batch import (
 from sglang.srt.mem_cache.memory_pool import TokenToKVPoolAllocator
 from sglang.srt.model_executor.forward_batch_info import CaptureHiddenMode
 from sglang.srt.speculative.build_eagle_tree import build_tree_kernel_efficient
-from sglang.srt.utils import fast_topk, is_cuda_available, is_hip, next_power_of_2
+from sglang.srt.utils import fast_topk, is_cuda, is_hip, next_power_of_2
 
-if is_cuda_available():
+if is_cuda():
     from sgl_kernel import (
         top_k_renorm_prob,
         top_p_renorm_prob,

python/sglang/srt/speculative/eagle_worker.py

@@ -34,14 +34,9 @@ from sglang.srt.speculative.eagle_utils import (
     select_top_k_tokens,
 )
 from sglang.srt.speculative.spec_info import SpeculativeAlgorithm
-from sglang.srt.utils import (
-    empty_context,
-    fast_topk,
-    get_available_gpu_memory,
-    is_cuda_available,
-)
+from sglang.srt.utils import empty_context, fast_topk, get_available_gpu_memory, is_cuda
 
-if is_cuda_available():
+if is_cuda():
     from sgl_kernel import segment_packbits
 
 logger = logging.getLogger(__name__)