[E2E] Optimize nightly testcase. (#4886)

### What this PR does / why we need it?
Optimize nightly testcase.
Changes:
- tests/e2e/nightly/multi_node/config/models/Qwen3-235B-A3B.yaml: Add
accuracy and performance benchmark
- tests/e2e/models/configs/Qwen3-8B-Base.yaml: Delete
- tests/e2e/models/configs/internlm-7b.yaml: Change to
internlm3-8b-instruct
- tests/e2e/nightly/models/test_deepseek_r1_w8a8_eplb.py: Change to
DeepSeek-R1-0528-W8A8 model

- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

Signed-off-by: menogrey <1299267905@qq.com>
This commit is contained in:
zhangyiming
2025-12-11 10:15:39 +08:00
committed by GitHub
parent 66b0781840
commit c95c271538
8 changed files with 17 additions and 20 deletions

View File

@@ -48,6 +48,17 @@ deployment:
--no-enable-prefix-caching
--gpu-memory-utilization 0.9
benchmarks:
perf:
case_type: performance
dataset_path: vllm-ascend/GSM8K-in3500-bs2800
request_conf: vllm_api_stream_chat
dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_str_perf
num_prompts: 2800
max_out_len: 1500
batch_size: 700
request_rate: 11.2
baseline: 1
threshold: 0.97
acc:
case_type: accuracy
dataset_path: vllm-ascend/gsm8k