chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Clean up imports and move files (#14317)

Browse Source
This commit is contained in:
Lianmin Zheng
2025-12-02 16:31:54 -08:00
committed by GitHub
parent 5c03aa3e9d
commit ca52ed425f
28 changed files with 280 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
python/sglang/srt/model_executor/cuda_graph_runner.py
View File

@@ -28,6 +28,7 @@ import torch
import tqdm
from torch.profiler import ProfilerActivity, profile
from sglang.srt.batch_overlap.two_batch_overlap import TboCudaGraphRunnerPlugin
from sglang.srt.constants import GPU_MEMORY_TYPE_CUDA_GRAPH
from sglang.srt.custom_op import CustomOp
from sglang.srt.distributed import get_tensor_model_parallel_rank
@@ -60,7 +61,6 @@ from sglang.srt.model_executor.forward_batch_info import (
)
from sglang.srt.model_executor.input_buffers import GraphInputBuffers
from sglang.srt.multiplex.pdmux_context import get_current_stream_idx, get_stream_groups
from sglang.srt.two_batch_overlap import TboCudaGraphRunnerPlugin
from sglang.srt.utils import (
empty_context,
get_available_gpu_memory,

2
python/sglang/srt/model_executor/forward_batch_info.py
View File

@@ -719,7 +719,7 @@ class ForwardBatch:
)
def prepare_mlp_sync_batch(self, model_runner: ModelRunner):
from sglang.srt.two_batch_overlap import TboForwardBatchPreparer
from sglang.srt.batch_overlap.two_batch_overlap import TboForwardBatchPreparer
assert self.global_num_tokens_cpu is not None
assert self.global_num_tokens_for_logprob_cpu is not None

2
python/sglang/srt/model_executor/piecewise_cuda_graph_runner.py
View File

@@ -24,6 +24,7 @@ from typing import TYPE_CHECKING, Union
import torch
import tqdm
from sglang.srt.batch_overlap.two_batch_overlap import TboCudaGraphRunnerPlugin
from sglang.srt.compilation.compilation_config import CompilationConfig
from sglang.srt.compilation.compile import install_torch_compiled, set_compiled
from sglang.srt.compilation.piecewise_context_manager import (
@@ -51,7 +52,6 @@ from sglang.srt.model_executor.forward_batch_info import (
ForwardMode,
PPProxyTensors,
)
from sglang.srt.two_batch_overlap import TboCudaGraphRunnerPlugin
from sglang.srt.utils import get_available_gpu_memory, log_info_on_rank0
logger = logging.getLogger(__name__)