Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
change
sglang
Commits
86e0dde5
You need to sign in or sign up before continuing.
Unverified
Commit
86e0dde5
authored
Oct 26, 2024
by
Lianmin Zheng
Committed by
GitHub
Oct 26, 2024
Browse files
Improve the user control of new_token_ratio (#1811)
parent
2b809788
Changes
4
Hide whitespace changes
Inline
Side-by-side
Showing
4 changed files
with
32 additions
and
17 deletions
+32
-17
python/sglang/global_config.py
python/sglang/global_config.py
+9
-3
python/sglang/srt/managers/scheduler.py
python/sglang/srt/managers/scheduler.py
+15
-6
python/sglang/srt/model_executor/cuda_graph_runner.py
python/sglang/srt/model_executor/cuda_graph_runner.py
+2
-2
python/sglang/srt/server_args.py
python/sglang/srt/server_args.py
+6
-6
No files found.
python/sglang/global_config.py
View file @
86e0dde5
...
...
@@ -14,9 +14,15 @@ class GlobalConfig:
self
.
default_backend
=
None
# Runtime constants: New generation token ratio estimation
self
.
init_new_token_ratio
=
0.7
self
.
base_min_new_token_ratio
=
0.1
self
.
new_token_ratio_decay
=
0.001
self
.
default_init_new_token_ratio
=
float
(
os
.
environ
.
get
(
"SGLANG_INIT_NEW_TOKEN_RATIO"
,
0.7
)
)
self
.
default_min_new_token_ratio_factor
=
float
(
os
.
environ
.
get
(
"SGLANG_MIN_NEW_TOKEN_RATIO_FACTOR"
,
0.14
)
)
self
.
default_new_token_ratio_decay_steps
=
float
(
os
.
environ
.
get
(
"SGLANG_NEW_TOKEN_RATIO_DECAY_STEPS"
,
600
)
)
# Runtime constants: others
self
.
retract_decode_steps
=
20
...
...
python/sglang/srt/managers/scheduler.py
View file @
86e0dde5
...
...
@@ -254,13 +254,22 @@ class Scheduler:
assert
(
server_args
.
schedule_conservativeness
>=
0
),
"Invalid schedule_conservativeness"
self
.
min_new_token_ratio
=
min
(
global_config
.
base_min_new_token_ratio
self
.
init_new_token_ratio
=
min
(
global_config
.
default_init_new_token_ratio
*
server_args
.
schedule_conservativeness
,
1.0
,
)
self
.
new_token_ratio
=
self
.
min_new_token_ratio
self
.
new_token_ratio_decay
=
global_config
.
new_token_ratio_decay
self
.
min_new_token_ratio
=
min
(
self
.
init_new_token_ratio
*
global_config
.
default_min_new_token_ratio_factor
,
1.0
,
)
self
.
new_token_ratio_decay
=
(
self
.
init_new_token_ratio
-
self
.
min_new_token_ratio
)
/
global_config
.
default_new_token_ratio_decay_steps
self
.
new_token_ratio
=
self
.
init_new_token_ratio
self
.
batch_is_full
=
False
# Init profiler
...
...
@@ -307,7 +316,7 @@ class Scheduler:
self
.
process_batch_result
(
batch
,
result
)
else
:
self
.
check_memory
()
self
.
new_token_ratio
=
global_config
.
init_new_token_ratio
self
.
new_token_ratio
=
self
.
init_new_token_ratio
self
.
last_batch
=
batch
...
...
@@ -334,7 +343,7 @@ class Scheduler:
self
.
process_batch_result
(
tmp_batch
,
tmp_result
)
elif
batch
is
None
:
self
.
check_memory
()
self
.
new_token_ratio
=
global_config
.
init_new_token_ratio
self
.
new_token_ratio
=
self
.
init_new_token_ratio
self
.
last_batch
=
batch
...
...
python/sglang/srt/model_executor/cuda_graph_runner.py
View file @
86e0dde5
...
...
@@ -121,13 +121,13 @@ class CudaGraphRunner:
bs
for
bs
in
self
.
capture_bs
if
bs
<=
model_runner
.
req_to_token_pool
.
size
and
bs
<=
model_runner
.
server_args
.
max_
cuda_graph_bs
and
bs
<=
model_runner
.
server_args
.
cuda_graph_
max_
bs
]
self
.
compile_bs
=
(
[
bs
for
bs
in
self
.
capture_bs
if
bs
<=
self
.
model_runner
.
server_args
.
max_
torch_compile_bs
if
bs
<=
self
.
model_runner
.
server_args
.
torch_compile_
max_
bs
]
if
self
.
use_torch_compile
else
[]
...
...
python/sglang/srt/server_args.py
View file @
86e0dde5
...
...
@@ -119,8 +119,8 @@ class ServerArgs:
enable_overlap_schedule
:
bool
=
False
enable_mixed_chunk
:
bool
=
False
enable_torch_compile
:
bool
=
False
max_
torch_compile_bs
:
int
=
32
max_
cuda_graph_bs
:
int
=
160
torch_compile_
max_
bs
:
int
=
32
cuda_graph_
max_
bs
:
int
=
160
torchao_config
:
str
=
""
enable_p2p_check
:
bool
=
False
triton_attention_reduce_in_fp32
:
bool
=
False
...
...
@@ -620,15 +620,15 @@ class ServerArgs:
help
=
"Optimize the model with torch.compile. Experimental feature."
,
)
parser
.
add_argument
(
"--
max-
torch-compile-bs"
,
"--torch-compile-
max-
bs"
,
type
=
int
,
default
=
ServerArgs
.
max_
torch_compile_bs
,
default
=
ServerArgs
.
torch_compile_
max_
bs
,
help
=
"Set the maximum batch size when using torch compile."
,
)
parser
.
add_argument
(
"--
max-
cuda-graph-bs"
,
"--cuda-graph-
max-
bs"
,
type
=
int
,
default
=
ServerArgs
.
max_
cuda_graph_bs
,
default
=
ServerArgs
.
cuda_graph_
max_
bs
,
help
=
"Set the maximum batch size for cuda graph."
,
)
parser
.
add_argument
(
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment