Files
enginex-ascend-910-vllm/tests/e2e/models/configs/gemma-3-4b-it.yaml
Sun Ruoxi 7f8a1b1f7a init v0.23.0
Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>
2026-08-27 15:11:51 +08:00

25 lines
472 B
YAML

model_name: "LLM-Research/gemma-3-4b-it"
model_type: "vllm"
hardware: "Atlas A2 Series"
serve:
tensor_parallel_size: 1
dtype: auto
max_model_len: 4096
gpu_memory_utilization: 0.7
trust_remote_code: false
enforce_eager: true
tasks:
- name: "gsm8k"
metrics:
- name: "exact_match,strict-match"
value: 0.59
- name: "exact_match,flexible-extract"
value: 0.59
num_fewshot: 5
apply_chat_template: false
fewshot_as_multiturn: false
batch_size: "auto"