Merge remote-tracking branch 'origin/big-refactor' into big-refactor_dp

lm_eval/tasks/glue/qnli/default.yaml

@@ -10,3 +10,5 @@ doc_to_target: label
 doc_to_choice: ["yes", "no"]
 metric_list:
   - metric: acc
+metadata:
+  - version: 1.0

lm_eval/tasks/glue/qqp/default.yaml

@@ -12,3 +12,5 @@ doc_to_choice: ["no", "yes"]
 metric_list:
   - metric: acc
   - metric: f1
+metadata:
+  - version: 1.0

lm_eval/tasks/glue/rte/default.yaml

@@ -10,3 +10,5 @@ doc_to_target: label
 doc_to_choice: ["True", "False"]
 metric_list:
   - metric: acc
+metadata:
+  - version: 1.0

lm_eval/tasks/glue/sst/default.yaml

@@ -10,3 +10,5 @@ doc_to_target: label
 doc_to_choice: ["negative", "positive"]
 metric_list:
   - metric: acc
+metadata:
+  - version: 1.0

lm_eval/tasks/glue/wnli/default.yaml

@@ -10,3 +10,5 @@ doc_to_target: label
 doc_to_choice: ["False", "True"]
 metric_list:
   - metric: acc
+metadata:
+  - version: 2.0

lm_eval/tasks/gsm8k/gsm8k-cot-self-consistency.yaml

@@ -30,3 +30,5 @@ filter_list:
         regex_pattern: "The answer is (\\-?[0-9\\.\\,]*[0-9]+)"
       - function: "majority_vote"
       - function: "take_first"
+metadata:
+  - version: 0.0

lm_eval/tasks/gsm8k/gsm8k-cot.yaml

@@ -40,3 +40,5 @@ filter_list:
       - function: "regex"
         regex_pattern: "The answer is (\\-?[0-9\\.\\,]+)."
       - function: "take_first"
+metadata:
+  - version: 0.0

lm_eval/tasks/gsm8k/gsm8k.yaml

@@ -33,3 +33,5 @@ filter_list:
       - function: "regex"
         regex_pattern: "#### (\\-?[0-9\\.\\,]+)"
       - function: "take_first"
+metadata:
+  - version: 1.0

lm_eval/tasks/headqa/headqa_en.yaml

@@ -19,3 +19,5 @@ metric_list:
   - metric: acc_norm
     aggregation: mean
     higher_is_better: true
+metadata:
+  - version: 1.0

lm_eval/tasks/hellaswag/hellaswag.yaml

@@ -18,3 +18,5 @@ metric_list:
   - metric: acc_norm
     aggregation: mean
     higher_is_better: true
+metadata:
+  - version: 1.0

lm_eval/tasks/hendrycks_ethics/commonsense.yaml

@@ -11,3 +11,5 @@ doc_to_target: label
 doc_to_choice: ['no', 'yes']
 metric_list:
   - metric: acc
+metadata:
+  - version: 1.0

lm_eval/tasks/hendrycks_ethics/deontology.yaml

@@ -4,5 +4,6 @@ dataset_name: deontology
 doc_to_text: "Question: Would most people believe this reasonable or unreasonable to say? \"{{scenario}} {{excuse.rstrip()}}\"\nAnswer:"
 doc_to_target: label
 doc_to_choice: ['unreasonable', 'reasonable']
-
+metadata:
+  - version: 1.0
 # TODO: implement exact-match metric for this subset

lm_eval/tasks/hendrycks_ethics/justice.yaml

@@ -5,3 +5,5 @@ task: ethics_justice
 dataset_name: justice
 doc_to_text: "Question: Would most people believe this reasonable or unreasonable to say? \"{{scenario}}\"\nAnswer:"
 # TODO: impl. exact match for this and deontology
+metadata:
+  - version: 1.0

lm_eval/tasks/hendrycks_ethics/utilitarianism.yaml

@@ -8,3 +8,5 @@ doc_to_target: !function utils.doc_to_target
 doc_to_choice: ['no', 'yes']
 metric_list:
   - metric: acc
+metadata:
+  - version: 1.0

lm_eval/tasks/hendrycks_ethics/utilitarianism_original_yaml

@@ -12,3 +12,5 @@
 # metric_list:
 #   - metric: acc
 # TODO: we want this to be implemented as a winograd_schema task type, actually
+# metadata:
+#   - version: 1.0

lm_eval/tasks/hendrycks_ethics/virtue.yaml

@@ -6,3 +6,5 @@ dataset_name: virtue
 doc_to_text: "Sentence: {{scenario}}\nQuestion: Does the character in this sentence exhibit the trait \"{{trait}}\"?\nAnswer:"
 doc_to_target: label
 doc_to_choice: ['no', 'yes']
+metadata:
+  - version: 1.0

lm_eval/tasks/lambada/lambada_openai.yaml

@@ -16,3 +16,5 @@ metric_list:
   - metric: acc
     aggregation: mean
     higher_is_better: true
+metadata:
+  - version: 1.0

lm_eval/tasks/lambada/lambada_standard.yaml

@@ -17,3 +17,5 @@ metric_list:
   - metric: acc
     aggregation: mean
     higher_is_better: true
+metadata:
+  - version: 1.0

lm_eval/tasks/lambada_cloze/lambada_openai_cloze.yaml

@@ -16,3 +16,5 @@ metric_list:
   - metric: acc
     aggregation: mean
     higher_is_better: true
+metadata:
+  - version: 1.0

lm_eval/tasks/lambada_cloze/lambada_standard_cloze.yaml

@@ -17,3 +17,5 @@ metric_list:
   - metric: acc
     aggregation: mean
     higher_is_better: true
+metadata:
+  - version: 1.0