chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

Revert several PRs (#14958)

Browse Source 
Co-authored-by: fzyzcjy <ch271828n@outlook.com>
This commit is contained in:
Yineng Zhang
2025-12-12 11:25:12 -08:00
committed by GitHub
co-authored by fzyzcjy
parent ec242f516e
commit 4b7b5af36a
11 changed files with 38 additions and 500 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/managers/scheduler.py
+7 -26
View File
@@ -2609,9 +2609,6 @@ class Scheduler(
self.send_to_detokenizer.send_output(recv_req, recv_req)
return None
def get_remote_instance_transfer_engine_info(self):
return self.tp_worker.get_remote_instance_transfer_engine_info()
class IdleSleeper:
"""
@@ -2725,29 +2722,13 @@ def run_scheduler_process(
pp_rank,
dp_rank,
)
if server_args.remote_instance_weight_loader_support_transfer_engine:
(
remote_instance_transfer_engine_session_id,
remote_instance_transfer_engine_weights_info_dict,
) = scheduler.get_remote_instance_transfer_engine_info()
pipe_writer.send(
{
"status": "ready",
"max_total_num_tokens": scheduler.max_total_num_tokens,
"max_req_input_len": scheduler.max_req_input_len,
"tp_rank": tp_rank,
"remote_instance_transfer_engine_session_id": remote_instance_transfer_engine_session_id,
"remote_instance_transfer_engine_weights_info_dict": remote_instance_transfer_engine_weights_info_dict,
}
)
else:
pipe_writer.send(
{
"status": "ready",
"max_total_num_tokens": scheduler.max_total_num_tokens,
"max_req_input_len": scheduler.max_req_input_len,
}
)
pipe_writer.send(
{
"status": "ready",
"max_total_num_tokens": scheduler.max_total_num_tokens,
"max_req_input_len": scheduler.max_req_input_len,
}
)
disaggregation_mode: DisaggregationMode = scheduler.disaggregation_mode
if disaggregation_mode == DisaggregationMode.NULL:
python/sglang/srt/managers/tp_worker.py
-6
View File
@@ -366,12 +366,6 @@ class TpModelWorker(BaseTpWorker):
can_run_cuda_graph=can_run_cuda_graph,
)
def get_remote_instance_transfer_engine_info(self):
return (
self.model_runner.remote_instance_transfer_engine_session_id,
self.model_runner.remote_instance_transfer_engine_weight_info,
)
def forward_batch_generation(
self,
model_worker_batch: ModelWorkerBatch,