Merge branch 'main' into africamgsm

lm_eval/tasks/afrimgsm/en_cot/cot_yaml

+# This file will be included in the generated language-specific task configs.
+# It doesn't have a yaml file extension as it is not meant to be imported directly
+# by the harness.
+group: afrimgsm_direct
+dataset_path: masakhane/afrimgsm
+dataset_name: null  # Overridden by language-specific config.
+output_type: generate_until
+training_split: train
+test_split: test
+generation_kwargs:
+  until:
+    - "\n\n"
+    - "\n"
+  do_sample: false
+  temperature: 0.0
+target_delimiter: " "
+metric_list:
+  - metric: exact_match
+    aggregation: mean
+    higher_is_better: true
+    ignore_case: true
+    ignore_punctuation: true
+filter_list:
+  - name: "strict-match"
+    filter:
+      - function: "regex"
+        regex_pattern: "The answer is (\\-?[0-9\\.\\,]+)"
+      - function: "take_first"
+  - filter:
+    - function: regex
+      group_select: -1
+      regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+    - function: take_first
+    name: flexible-extract
+metadata:
+  version: 2.0

lm_eval/tasks/afrimgsm/gen_yaml.sh

 #!/bin/bash
 
-python utils.py --overwrite --output-dir direct --mode direct
+# python utils.py --overwrite --output-dir direct --mode direct
+# python utils.py --overwrite --output-dir direct_native --mode direct-native
 # python utils.py --overwrite --output-dir en_cot --mode en-cot
 # python utils.py --overwrite --output-dir native_cot --mode native-cot
+python utils.py --overwrite --output-dir translate_direct --mode translate-direct

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_amh.yaml

+# Generated by utils.py
+dataset_name: amh
+doc_to_target: '{% if answer is not none %}{{answer[15:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nበቅደም ተከተል መልስ:"}}{% else %}{{"ጥያቄ: "+question+"\nበቅደም ተከተል መልስ:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: መልሱ (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'ጥያቄ:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_amh

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_eng.yaml

+# Generated by utils.py
+dataset_name: eng
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: The answer is (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_eng

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_ewe.yaml

+# Generated by utils.py
+dataset_name: ewe
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: ŋuɖoɖoae nye (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_ewe

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_fra.yaml

+# Generated by utils.py
+dataset_name: fra
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: La réponse est(\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_fra

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_hau.yaml

+# Generated by utils.py
+dataset_name: hau
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Amsar ita ce (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_hau

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_ibo.yaml

+# Generated by utils.py
+dataset_name: ibo
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Azịza ya bụ (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_ibo

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_kin.yaml

+# Generated by utils.py
+dataset_name: kin
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Igisubizo ni (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_kin

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_lin.yaml

+# Generated by utils.py
+dataset_name: lin
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Eyano ezali (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_lin

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_lug.yaml

+# Generated by utils.py
+dataset_name: lug
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Ansa eri (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_lug

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_orm.yaml

+# Generated by utils.py
+dataset_name: orm
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Deebiin isaa (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_orm

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_sna.yaml

+# Generated by utils.py
+dataset_name: sna
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Mhinduro kumubvunzo ndi (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_sna

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_sot.yaml

+# Generated by utils.py
+dataset_name: sot
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Karabo ke (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_sot

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_swa.yaml

+# Generated by utils.py
+dataset_name: swa
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Jibu ni (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_swa

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_twi.yaml

+# Generated by utils.py
+dataset_name: twi
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Ne nnyiano yɛ (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_twi

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_wol.yaml

+# Generated by utils.py
+dataset_name: wol
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Tontu li (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_wol

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_xho.yaml

+# Generated by utils.py
+dataset_name: xho
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Impendulo ngu (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_xho

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_yor.yaml

+# Generated by utils.py
+dataset_name: yor
+doc_to_target: '{% if answer is not none %}{{answer[16:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nÌdáhùn lẹ́sẹsẹ:"}}{% else %}{{"Ìbéèrè: "+question+"\nÌdáhùn lẹ́sẹsẹ:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Ìdáhùn náà ni (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Ìbéèrè:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_yor

lm_eval/tasks/afrimgsm/native_cot/afrimgsm_native_cot_zul.yaml

+# Generated by utils.py
+dataset_name: zul
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+filter_list:
+- filter:
+  - function: regex
+    regex_pattern: Impendulo ithi (\-?[0-9\.\,]+)
+  - function: take_first
+  name: strict-match
+- filter:
+  - function: regex
+    group_select: -1
+    regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+  - function: take_first
+  name: flexible-extract
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_native_cot_zul