|
1
|
model_name: "RedHatAI/Qwen3.5-4B-quantized.w4a16" |
|
2
|
accuracy_threshold: 0.82 |
|
3
|
num_questions: 1319 |
|
4
|
num_fewshot: 5 |
|
5
|
gen_prefix: " <think>\n\n</think>\n" |
|
6
|
server_args: >- |
|
7
|
--enforce-eager |
|
8
|
--max-model-len 8192 |
|
9
|
env: |
|
10
|
VLLM_HUMMING_INPUT_QUANT_CONFIG: '{"dtype":"float8e4m3"}' |