[Test] Add accuracy nightly test for new models (#4262)

### What this PR does / why we need it?
Add accuracy nightly test for new models:

PaddlePaddle/ERNIE-4.5-21B-A3B-PT
LLM-Research/Molmo-7B-D-0924
LLM-Research/gemma-2-9b-it
LLM-Research/gemma-3-4b-it
Shanghai_AI_Laboratory/internlm-7b
llava-hf/llava-1.5-7b-hf

- vLLM version: v0.11.2

Signed-off-by: hfadzxy <starmoon_zhang@163.com>
This commit is contained in:
zhangxinyuehfad
2025-12-01 22:28:46 +08:00
committed by GitHub
parent 8e7f5cff6d
commit b6afec73e1
11 changed files with 97 additions and 4 deletions

View File

@@ -0,0 +1,13 @@
model_name: "LLM-Research/Molmo-7B-D-0924"
hardware: "Atlas A2 Series"
model: "vllm-vlm"
tasks:
- name: "ceval-valid"
metrics:
- name: "acc,none"
value: 0.71
max_model_len: 4096
trust_remote_code: True
apply_chat_template: False
fewshot_as_multiturn: False
gpu_memory_utilization: 0.8