[test] add npu test yaml and add ascend a3 docker file (#9547)

Co-authored-by: jiaqiw09 <jiaqiw960714@gmail.com>
This commit is contained in:
Username_Full
2025-11-30 09:37:08 +08:00
committed by GitHub
parent 22be45c78c
commit e43a972b25
33 changed files with 322 additions and 21 deletions

View File

@@ -14,6 +14,8 @@
import os
import pytest
from llamafactory.chat import ChatModel
@@ -35,11 +37,13 @@ MESSAGES = [
EXPECTED_RESPONSE = "_rho"
@pytest.mark.runs_on(["cpu"])
def test_chat():
chat_model = ChatModel(INFER_ARGS)
assert chat_model.chat(MESSAGES)[0].response_text == EXPECTED_RESPONSE
@pytest.mark.runs_on(["cpu"])
def test_stream_chat():
chat_model = ChatModel(INFER_ARGS)
response = ""

View File

@@ -39,6 +39,7 @@ MESSAGES = [
]
@pytest.mark.runs_on(["cpu"])
@pytest.mark.skipif(not is_sglang_available(), reason="SGLang is not installed")
def test_chat():
r"""Test the SGLang engine's basic chat functionality."""
@@ -48,6 +49,7 @@ def test_chat():
print(response.response_text)
@pytest.mark.runs_on(["cpu"])
@pytest.mark.skipif(not is_sglang_available(), reason="SGLang is not installed")
def test_stream_chat():
r"""Test the SGLang engine's streaming chat functionality."""

View File

@@ -48,7 +48,7 @@ INFER_ARGS = {
OS_NAME = os.getenv("OS_NAME", "")
@pytest.mark.runs_on(["cpu"])
@pytest.mark.parametrize(
"stage,dataset",
[
@@ -65,6 +65,7 @@ def test_run_exp(stage: str, dataset: str):
assert os.path.exists(output_dir)
@pytest.mark.runs_on(["cpu"])
def test_export():
export_dir = os.path.join("output", "llama3_export")
export_model({"export_dir": export_dir, **INFER_ARGS})