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
b8ccaf4d
Unverified
Commit
b8ccaf4d
authored
Sep 21, 2024
by
Ke Bao
Committed by
GitHub
Sep 21, 2024
Browse files
Add MLA gsm8k eval (#1484)
parent
a68cb201
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
15 additions
and
3 deletions
+15
-3
.github/workflows/pr-test.yml
.github/workflows/pr-test.yml
+1
-1
python/sglang/test/test_utils.py
python/sglang/test/test_utils.py
+2
-2
test/srt/test_mla.py
test/srt/test_mla.py
+12
-0
No files found.
.github/workflows/pr-test.yml
View file @
b8ccaf4d
...
...
@@ -238,7 +238,7 @@ jobs:
timeout-minutes
:
10
run
:
|
cd test/srt
python3
-m unittest test_mla.TestMLA.test_mmlu
python3
test_mla.py
finish
:
needs
:
[
...
...
python/sglang/test/test_utils.py
View file @
b8ccaf4d
...
...
@@ -28,9 +28,9 @@ 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_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"
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
"
DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_FP8_TP1
=
"neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8,neuralmagic/Mistral-7B-Instruct-v0.3-FP8,neuralmagic/DeepSeek-Coder-V2-Lite-Instruct-FP8,neuralmagic/gemma-2-2b-it-FP8"
DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_FP8_TP2
=
"neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8,neuralmagic/Mixtral-8x7B-Instruct-v0.1-FP8,neuralmagic/Qwen2-72B-Instruct-FP8,neuralmagic/Qwen2-57B-A14B-Instruct-FP8"
DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_FP8_TP2
=
"neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8,neuralmagic/Mixtral-8x7B-Instruct-v0.1-FP8,neuralmagic/Qwen2-72B-Instruct-FP8,neuralmagic/Qwen2-57B-A14B-Instruct-FP8
,neuralmagic/DeepSeek-Coder-V2-Lite-Instruct-FP8
"
DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_QUANT_TP1
=
"hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4,hugging-quants/Meta-Llama-3.1-8B-Instruct-GPTQ-INT4"
...
...
test/srt/test_mla.py
View file @
b8ccaf4d
...
...
@@ -39,6 +39,18 @@ class TestMLA(unittest.TestCase):
metrics
=
run_eval
(
args
)
assert
metrics
[
"score"
]
>=
0.5
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