Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
gaoqiong
lm-evaluation-harness
Commits
c0d5a660
Commit
c0d5a660
authored
Jan 17, 2024
by
lintangsutawika
Browse files
resolved conflict
parents
f7b81bd4
0d1ef037
Changes
1000
Hide whitespace changes
Inline
Side-by-side
Showing
20 changed files
with
162 additions
and
0 deletions
+162
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_3ds.yaml
...rithmetic/alternative_worlds/style_04/arithmetic_3ds.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_4da.yaml
...rithmetic/alternative_worlds/style_04/arithmetic_4da.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_4ds.yaml
...rithmetic/alternative_worlds/style_04/arithmetic_4ds.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_5da.yaml
...rithmetic/alternative_worlds/style_04/arithmetic_5da.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_5ds.yaml
...rithmetic/alternative_worlds/style_04/arithmetic_5ds.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/_template_05_yaml
.../arithmetic/alternative_worlds/style_05/_template_05_yaml
+15
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_1dc.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_1dc.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_2da.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_2da.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_2dm.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_2dm.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_2ds.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_2ds.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_3da.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_3da.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_3ds.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_3ds.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_4da.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_4da.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_4ds.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_4ds.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_5da.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_5da.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_5ds.yaml
...rithmetic/alternative_worlds/style_05/arithmetic_5ds.yaml
+4
-0
lm_eval/tasks/arithmetic/alternative_worlds/utils.py
lm_eval/tasks/arithmetic/alternative_worlds/utils.py
+36
-0
lm_eval/tasks/bbh/alternative_worlds/README.md
lm_eval/tasks/bbh/alternative_worlds/README.md
+34
-0
lm_eval/tasks/bbh/alternative_worlds/prompt_variation/bbh_alt_pv_zeroshot.yaml
...ernative_worlds/prompt_variation/bbh_alt_pv_zeroshot.yaml
+5
-0
lm_eval/tasks/bbh/alternative_worlds/prompt_variation/style_01/zeroshot/_zeroshot_template_yaml
...rompt_variation/style_01/zeroshot/_zeroshot_template_yaml
+12
-0
No files found.
Too many changes to show.
To preserve performance only
1000 of 1000+
files are displayed.
Plain diff
Email patch
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_3ds.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_04_yaml
task
:
arithmetic_3ds_alt_04
dataset_name
:
arithmetic_3ds
task_alias
:
3ds
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_4da.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_04_yaml
task
:
arithmetic_4da_alt_04
dataset_name
:
arithmetic_4da
task_alias
:
4da
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_4ds.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_04_yaml
task
:
arithmetic_4ds_alt_04
dataset_name
:
arithmetic_4ds
task_alias
:
4ds
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_5da.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_04_yaml
task
:
arithmetic_5da_alt_04
dataset_name
:
arithmetic_5da
task_alias
:
5da
lm_eval/tasks/arithmetic/alternative_worlds/style_04/arithmetic_5ds.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_04_yaml
task
:
arithmetic_5ds_alt_04
dataset_name
:
arithmetic_5ds
task_alias
:
5ds
lm_eval/tasks/arithmetic/alternative_worlds/style_05/_template_05_yaml
0 → 100644
View file @
c0d5a660
include: ../_template_yaml
group: arithmetic_alt_05
group_alias: arithmetic (Style 05)
dataset_path: EleutherAI/arithmetic
output_type: loglikelihood
validation_split: validation
test_split: null
doc_to_text: !function ../utils.style_05
doc_to_target: "{{completion}}"
metric_list:
- metric: acc
aggregation: mean
higher_is_better: true
- metric: brier_score
higher_is_better: false
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_1dc.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_1dc_alt_05
dataset_name
:
arithmetic_1dc
task_alias
:
1dc
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_2da.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_2da_alt_05
dataset_name
:
arithmetic_2da
task_alias
:
2da
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_2dm.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_2dm_alt_05
dataset_name
:
arithmetic_2dm
task_alias
:
2dm
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_2ds.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_2ds_alt_05
dataset_name
:
arithmetic_2ds
task_alias
:
2ds
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_3da.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_3da_alt_05
dataset_name
:
arithmetic_3da
task_alias
:
3da
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_3ds.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_3ds_alt_05
dataset_name
:
arithmetic_3ds
task_alias
:
3ds
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_4da.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_4da_alt_05
dataset_name
:
arithmetic_4da
task_alias
:
4da
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_4ds.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_4ds_alt_05
dataset_name
:
arithmetic_4ds
task_alias
:
4ds
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_5da.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_5da_alt_05
dataset_name
:
arithmetic_5da
task_alias
:
5da
lm_eval/tasks/arithmetic/alternative_worlds/style_05/arithmetic_5ds.yaml
0 → 100644
View file @
c0d5a660
include
:
_template_05_yaml
task
:
arithmetic_5ds_alt_05
dataset_name
:
arithmetic_5ds
task_alias
:
5ds
lm_eval/tasks/arithmetic/alternative_worlds/utils.py
0 → 100644
View file @
c0d5a660
import
re
# Original Prompt
# Question: What is (9 + 8) * 2? Answer:
def
style_00
(
docs
):
# What is (9 + 8) * 2?
return
docs
[
"context"
]
def
style_01
(
docs
):
# What is (9 + 8) * 2?
return
docs
[
"context"
].
replace
(
"Question: "
,
""
).
replace
(
" Answer:"
,
""
)
def
style_02
(
docs
):
# Q: What is (9 + 8) * 2? A:
return
docs
[
"context"
].
replace
(
"Question: "
,
"Q: "
).
replace
(
" Answer:"
,
" A:"
)
def
style_03
(
docs
):
# Solve (9 + 8) * 2.
return
(
docs
[
"context"
].
replace
(
"Question: What is"
,
"Solve"
).
replace
(
" Answer:"
,
"."
)
)
def
style_04
(
docs
):
# (9 + 8) * 2 =
return
docs
[
"context"
].
replace
(
"Question: What is "
,
""
).
replace
(
" Answer:"
,
" ="
)
def
style_05
(
docs
):
# What is (9 + 8) * 2? Answer:
return
docs
[
"context"
].
replace
(
"Question: "
,
""
)
lm_eval/tasks/bbh/alternative_worlds/README.md
0 → 100644
View file @
c0d5a660
| Task | Prompt Variation | Output Variation | Option in Sample |
| :-----------------:| :---------------: | :---------------: |:---------------: |
| boolean_expression | Yes | Yes | No |
| causal_judgement | Yes | Yes | Yes |
| date_understanding | Yes | Yes | Yes |
| disambiguation_qa | Yes | Yes | Yes |
| dyck_languages | Yes | No | No |
| formal_fallacies | Yes | Yes | Yes |
| geometric_shapes | Yes | Yes | Yes |
| hyperbaton | Yes | Yes | Yes |
| logical_deduction_five_objects| Yes | Yes | Yes |
| logical_deduction_seven_objects| Yes | Yes | Yes |
| logical_deduction_three_objects| Yes | Yes | Yes |
| movie_recommendation| Yes | Yes | Yes |
| multistep_arithmetic_two| Yes | No | No |
| navigate | Yes | Yes | Yes |
| object_counting | Yes | No | No |
| penguins_in_a_table| Yes | Yes | Yes |
| reasoning_about_colored_objects| Yes | Yes | Yes |
| ruin_names | Yes | Yes | Yes |
| salient_translation_error_detection| Yes| Yes | Yes |
| snarks | Yes | Yes | Yes |
| sports_understanding| Yes | Yes | No |
| temporal_sequences | Yes | Yes | Yes |
| tracking_shuffled_objects_five_objects| Yes| Yes | Yes |
| tracking_shuffled_objects_seven_objects| Yes| Yes | Yes |
| tracking_shuffled_objects_three_objects| Yes| Yes | Yes |
| web_of_lies | Yes | Yes | No |
| word_sorting | Yes | No | No |
Notes:
-
`web_of_lies`
already starts with
`Question: `
-
Tasks with options are
`Options: (A) ...`
(multiple choice) or
`Options: - ...`
(binary choice)
lm_eval/tasks/bbh/alternative_worlds/prompt_variation/bbh_alt_pv_zeroshot.yaml
0 → 100644
View file @
c0d5a660
group
:
bbh_alt_pv_zeroshot
task
:
-
bbh_alt_pv_01_zeroshot
-
bbh_alt_pv_02_zeroshot
-
bbh_alt_pv_03_zeroshot
lm_eval/tasks/bbh/alternative_worlds/prompt_variation/style_01/zeroshot/_zeroshot_template_yaml
0 → 100644
View file @
c0d5a660
group: bbh_alt_pv_01_zeroshot
dataset_path: lukaemon/bbh
output_type: multiple_choice
test_split: test
doc_to_text: !function ../../styles.styles_01
doc_to_target: !function ../../styles.doc_to_target
doc_to_choice: !function ../../styles.doc_to_choice
num_fewshot: 0
metric_list:
- metric: acc
- metric: acc_norm
- metric: brier_score
Prev
1
…
3
4
5
6
7
8
9
10
11
…
50
Next
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment