Upload folder using huggingface_hub
Browse files- evals/tasks/mmlu_en_us.yaml +25 -0
evals/tasks/mmlu_en_us.yaml
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
task: mmlu_all
|
| 3 |
+
dataset_path: cais/mmlu
|
| 4 |
+
dataset_name: all
|
| 5 |
+
description: "MMLU dataset"
|
| 6 |
+
test_split: test
|
| 7 |
+
fewshot_split: dev
|
| 8 |
+
fewshot_config:
|
| 9 |
+
sampler: first_n
|
| 10 |
+
output_type: multiple_choice
|
| 11 |
+
doc_to_text: "{{question.strip()}}
|
| 12 |
+
A. {{choices[0]}}
|
| 13 |
+
B. {{choices[1]}}
|
| 14 |
+
C. {{choices[2]}}
|
| 15 |
+
D. {{choices[3]}}
|
| 16 |
+
Answer:"
|
| 17 |
+
doc_to_choice: ["A", "B", "C", "D"]
|
| 18 |
+
doc_to_target: answer
|
| 19 |
+
metric_list:
|
| 20 |
+
- metric: acc
|
| 21 |
+
aggregation: mean
|
| 22 |
+
higher_is_better: true
|
| 23 |
+
- metric: acc_norm
|
| 24 |
+
aggregation: mean
|
| 25 |
+
higher_is_better: true
|