A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 11, 16:15 UTC
[Bugfix] Fix `task` still being passed in tests/benchmarks (#30476)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Dec 11, 2025 at 18:33 UTC
d917747c95b212f9b7e85c100bc572e3e5d33360
2 files changed
-3
benchmarks/benchmark_ngram_proposer.py
-1
@@ -32,7 +32,6 @@ def benchmark_propose(args):
32
33
model_config = ModelConfig(
34
model="facebook/opt-125m",
35
- task="generate",
35
max_model_len=args.num_token + args.num_spec_token,
36
tokenizer="facebook/opt-125m",
37
tokenizer_mode="auto",
tests/models/language/pooling/test_mm_classifier_conversion.py
-2
@@ -17,7 +17,6 @@ def test_idefics_multimodal(
17
with vllm_runner(
18
model_name="HuggingFaceM4/Idefics3-8B-Llama3",
19
runner="pooling",
20
- task="classify",
20
convert="classify",
21
load_format="dummy",
22
max_model_len=512,
@@ -86,7 +85,6 @@ def test_gemma_multimodal(
85
with vllm_runner(
86
model_name="google/gemma-3-4b-it",
87
runner="pooling",
89
- task="classify",
88
convert="classify",
89
load_format="auto",
90
hf_overrides=update_config,