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
42a2d82b
Unverified
Commit
42a2d82b
authored
Sep 23, 2024
by
Yineng Zhang
Committed by
GitHub
Sep 23, 2024
Browse files
minor: add mla fp8 test (#1494)
parent
e4780cf8
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
52 additions
and
0 deletions
+52
-0
.github/workflows/pr-test.yml
.github/workflows/pr-test.yml
+1
-0
python/sglang/test/test_utils.py
python/sglang/test/test_utils.py
+1
-0
test/srt/test_mla_fp8.py
test/srt/test_mla_fp8.py
+50
-0
No files found.
.github/workflows/pr-test.yml
View file @
42a2d82b
...
...
@@ -239,6 +239,7 @@ jobs:
run
:
|
cd test/srt
python3 test_mla.py
python3 test_mla_fp8.py
-
name
:
Evaluate Data Parallelism Accuracy (TP=2)
timeout-minutes
:
10
...
...
python/sglang/test/test_utils.py
View file @
42a2d82b
...
...
@@ -26,6 +26,7 @@ DEFAULT_FP8_MODEL_NAME_FOR_TEST = "neuralmagic/Meta-Llama-3.1-8B-FP8"
DEFAULT_MODEL_NAME_FOR_TEST
=
"meta-llama/Meta-Llama-3.1-8B-Instruct"
DEFAULT_MOE_MODEL_NAME_FOR_TEST
=
"mistralai/Mixtral-8x7B-Instruct-v0.1"
DEFAULT_MLA_MODEL_NAME_FOR_TEST
=
"deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct"
DEFAULT_MLA_FP8_MODEL_NAME_FOR_TEST
=
"neuralmagic/DeepSeek-Coder-V2-Lite-Instruct-FP8"
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
=
600
DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_TP1
=
"meta-llama/Meta-Llama-3.1-8B-Instruct,mistralai/Mistral-7B-Instruct-v0.3,deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct,google/gemma-2-27b-it"
DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_TP2
=
"meta-llama/Meta-Llama-3.1-70B-Instruct,mistralai/Mixtral-8x7B-Instruct-v0.1,Qwen/Qwen2-57B-A14B-Instruct,deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct"
...
...
test/srt/test_mla_fp8.py
0 → 100644
View file @
42a2d82b
import
unittest
from
types
import
SimpleNamespace
from
sglang.srt.utils
import
kill_child_process
from
sglang.test.run_eval
import
run_eval
from
sglang.test.test_utils
import
(
DEFAULT_MLA_FP8_MODEL_NAME_FOR_TEST
,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
,
DEFAULT_URL_FOR_TEST
,
popen_launch_server
,
)
class
TestMLA
(
unittest
.
TestCase
):
@
classmethod
def
setUpClass
(
cls
):
cls
.
model
=
DEFAULT_MLA_FP8_MODEL_NAME_FOR_TEST
cls
.
base_url
=
DEFAULT_URL_FOR_TEST
cls
.
process
=
popen_launch_server
(
cls
.
model
,
cls
.
base_url
,
timeout
=
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
,
other_args
=
[
"--tp"
,
"2"
,
"--trust-remote-code"
,
"--kv-cache-dtype"
,
"fp8_e5m2"
,
],
)
@
classmethod
def
tearDownClass
(
cls
):
kill_child_process
(
cls
.
process
.
pid
)
def
test_mgsm_en
(
self
):
args
=
SimpleNamespace
(
base_url
=
self
.
base_url
,
model
=
self
.
model
,
eval_name
=
"mgsm_en"
,
num_examples
=
None
,
num_threads
=
1024
,
)
metrics
=
run_eval
(
args
)
assert
metrics
[
"score"
]
>=
0.8
if
__name__
==
"__main__"
:
unittest
.
main
()
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