biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
tests/models/quantization/test_nvfp4.py
View File

@@ -4,6 +4,7 @@
"""Tests Model Optimizer nvfp4 models against ground truth generation
Note: these tests will only pass on B200
"""
import os
from typing import List
@@ -21,14 +22,14 @@ MODELS = ["nvidia/Llama-3.3-70B-Instruct-FP4"]
EXPECTED_STRS_MAP = {
"nvidia/Llama-3.3-70B-Instruct-FP4": [
'vLLM (Vectorized Large Language Model) is indeed a high-throughput and memory-efficient inference',
'Here are the major milestones in the development of artificial intelligence (AI) from 1950 to ',
'Artificial intelligence (AI) and human intelligence (HI) are two distinct forms of intelligence that process',
'A neural network is a type of machine learning model inspired by the structure and function of the human brain',
'In the heart of a cutting-edge robotics lab, a team of engineers had been working tirelessly to push',
'The COVID-19 pandemic has had a profound impact on global economic structures and future business models, leading',
'The Mona Lisa, painted by Leonardo da Vinci in the early 16th century, is one of',
'Here are the translations:\n\n* Japanese: (Sasuga no tori ga miwa o ts'
"vLLM (Vectorized Large Language Model) is indeed a high-throughput and memory-efficient inference",
"Here are the major milestones in the development of artificial intelligence (AI) from 1950 to ",
"Artificial intelligence (AI) and human intelligence (HI) are two distinct forms of intelligence that process",
"A neural network is a type of machine learning model inspired by the structure and function of the human brain",
"In the heart of a cutting-edge robotics lab, a team of engineers had been working tirelessly to push",
"The COVID-19 pandemic has had a profound impact on global economic structures and future business models, leading",
"The Mona Lisa, painted by Leonardo da Vinci in the early 16th century, is one of",
"Here are the translations:\n\n* Japanese: (Sasuga no tori ga miwa o ts",
]
}
@@ -39,11 +40,13 @@ EXPECTED_STRS_MAP = {
# the hardware being run on.
# Disabled to prevent it from breaking the build
@pytest.mark.skip(
reason=
"Prevent unstable test based on golden strings from breaking the build "
" and test input model being too large and hanging the system.")
@pytest.mark.skipif(not is_quant_method_supported("modelopt_fp4"),
reason="modelopt_fp4 is not supported on this GPU type.")
reason="Prevent unstable test based on golden strings from breaking the build "
" and test input model being too large and hanging the system."
)
@pytest.mark.skipif(
not is_quant_method_supported("modelopt_fp4"),
reason="modelopt_fp4 is not supported on this GPU type.",
)
@pytest.mark.parametrize("model_name", MODELS)
def test_models(example_prompts, model_name) -> None:
llm = LLM(
@@ -56,12 +59,11 @@ def test_models(example_prompts, model_name) -> None:
tokenizer = AutoTokenizer.from_pretrained(model_name)
formatted_prompts = [
tokenizer.apply_chat_template([{
"role": "user",
"content": prompt
}],
tokenize=False,
add_generation_prompt=True)
tokenizer.apply_chat_template(
[{"role": "user", "content": prompt}],
tokenize=False,
add_generation_prompt=True,
)
for prompt in example_prompts
]
params = SamplingParams(max_tokens=20, temperature=0)
@@ -79,4 +81,5 @@ def test_models(example_prompts, model_name) -> None:
generated_str = generations[i]
expected_str = expected_strs[i]
assert expected_str == generated_str, (
f"Test{i}:\nExpected: {expected_str!r}\nvLLM: {generated_str!r}")
f"Test{i}:\nExpected: {expected_str!r}\nvLLM: {generated_str!r}"
)