Unverified Commit ce216c80 authored by Qubitium's avatar Qubitium Committed by GitHub
Browse files

Cleanup codebase: removed unnecessary code/logic (#298)

parent 51104cd4
...@@ -41,11 +41,15 @@ class RouterManager: ...@@ -41,11 +41,15 @@ class RouterManager:
self.send_to_detokenizer.send_pyobj(obj) self.send_to_detokenizer.send_pyobj(obj)
# async sleep for receiving the subsequent request and avoiding cache miss # async sleep for receiving the subsequent request and avoiding cache miss
slept = False
if len(out_pyobjs) != 0: if len(out_pyobjs) != 0:
has_finished = any([obj.finished for obj in out_pyobjs]) has_finished = any([obj.finished for obj in out_pyobjs])
if has_finished: if has_finished:
if self.extend_dependency_time > 0:
slept = True
await asyncio.sleep(self.extend_dependency_time) await asyncio.sleep(self.extend_dependency_time)
if not slept:
await asyncio.sleep(0.0006) await asyncio.sleep(0.0006)
async def loop_for_recv_requests(self): async def loop_for_recv_requests(self):
......
...@@ -90,7 +90,6 @@ class ModelRpcServer(rpyc.Service): ...@@ -90,7 +90,6 @@ class ModelRpcServer(rpyc.Service):
tokenizer_mode=server_args.tokenizer_mode, tokenizer_mode=server_args.tokenizer_mode,
trust_remote_code=server_args.trust_remote_code, trust_remote_code=server_args.trust_remote_code,
) )
self.eos_token_id = self.tokenizer.eos_token_id
self.max_total_num_token = self.model_runner.max_total_num_token self.max_total_num_token = self.model_runner.max_total_num_token
self.max_num_running_seq = self.max_total_num_token // 2 self.max_num_running_seq = self.max_total_num_token // 2
self.max_prefill_num_token = max( self.max_prefill_num_token = max(
......
...@@ -36,7 +36,6 @@ class ReqState: ...@@ -36,7 +36,6 @@ class ReqState:
out_list: List out_list: List
finished: bool finished: bool
event: asyncio.Event event: asyncio.Event
lock: asyncio.Lock
global global_processor global global_processor
...@@ -178,9 +177,8 @@ class TokenizerManager: ...@@ -178,9 +177,8 @@ class TokenizerManager:
) )
self.send_to_router.send_pyobj(tokenized_obj) self.send_to_router.send_pyobj(tokenized_obj)
lock = asyncio.Lock()
event = asyncio.Event() event = asyncio.Event()
state = ReqState([], False, event, lock) state = ReqState([], False, event)
self.rid_to_state[rid] = state self.rid_to_state[rid] = state
while True: while True:
...@@ -221,9 +219,8 @@ class TokenizerManager: ...@@ -221,9 +219,8 @@ class TokenizerManager:
) )
self.send_to_router.send_pyobj(tokenized_obj) self.send_to_router.send_pyobj(tokenized_obj)
lock = asyncio.Lock()
event = asyncio.Event() event = asyncio.Event()
state = ReqState([], False, event, lock) state = ReqState([], False, event)
self.rid_to_state[rid] = state self.rid_to_state[rid] = state
output_list = [] output_list = []
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment