prototype JudgeTask

lm_eval/api/judge_task.py

+import json
+
+import datasets
+
+from lm_eval.api.task import ConfigurableTask
+
+
+class JudgeTask(ConfigurableTask):
+    def __init__(self, config, output_path):
+        super().__init__(config)
+        self.output_path = output_path
+
+    def process_docs(self, dataset: datasets.Dataset):
+        resps = []
+        # load json
+        with open(self.output_path, "r") as f:
+            resp = json.load(f)
+            resps.append({"resp": resp["resps"], "doc": resp["doc"]})
+
+        resps.sort(key=lambda x: x["doc"])
+        dataset.add_column("resp", resps)
+        return resps

lm_eval/tasks/__init__.py

@@ -7,6 +7,7 @@ from typing import Dict, List, Mapping, Optional, Union
 
 from lm_eval import utils
 from lm_eval.api.group import ConfigurableGroup, GroupConfig
+from lm_eval.api.judge_task import JudgeTask
 from lm_eval.api.task import ConfigurableTask, Task
 from lm_eval.evaluator_utils import get_subtask_list
 
@@ -273,7 +274,12 @@ class TaskManager:
                     # very scuffed: set task name here. TODO: fixme?
                     task_object.config.task = config["task"]
             else:
-                task_object = ConfigurableTask(config=config)
+                if "resp_to_doc" in config:
+                    task_object = JudgeTask(
+                        config=config, output_path=config.get("output_path", None)
+                    )
+                else:
+                    task_object = ConfigurableTask(config=config)
 
             return {task: task_object}
 

lm_eval/tasks/judge/gsm8k_judge.yaml

+tag:
+  - judge
+include: gsm8k_judge_1.yaml
+task: gsm8k_judge_2
+output_type: generate_until
+
+output_path:
+
+doc_to_text: "Question: {{question}}\nAnswer:"
+doc_to_target: "{{answer}}" #" {{answer.split('### ')[-1].rstrip()}}"
+metric_list:
+  - metric: exact_match
+    aggregation: mean
+    higher_is_better: true
+    ignore_case: true
+    ignore_punctuation: false
+    regexes_to_ignore:
+      - ","
+      - "\\$"
+      - "(?s).*#### "
+      - "\\.$"
+generation_kwargs:
+  until:
+    - "Question:"
+    - "</s>"
+    - "<|im_end|>"
+  do_sample: false
+  temperature: 0.0
+repeats: 1
+num_fewshot: 5
+filter_list:
+  - name: "strict-match"
+    filter:
+      - function: "regex"
+        regex_pattern: "#### (\\-?[0-9\\.\\,]+)"
+      - function: "take_first"
+  - name: "flexible-extract"
+    filter:
+      - function: "regex"
+        group_select: -1
+        regex_pattern: "(-?[$0-9.,]{2,})|(-?[0-9]+)"
+      - function: "take_first"
+metadata:
+  version: 3.0

lm_eval/tasks/judge/gsm8k_judge_1.yaml

+tag:
+  - math_word_problems
+task: gsm8k
+dataset_path: gsm8k
+dataset_name: main
+output_type: generate_until
+training_split: train
+fewshot_split: train
+test_split: test
+doc_to_text: "Solve the following math problem step by step. The last line of your response should be of the form Answer: $ANSWER (without quotes) where $ANSWER is the answer to the problem\n\nQuestion: {{question}}\nAnswer:"
+doc_to_target: "{{answer}}" #" {{answer.split('### ')[-1].rstrip()}}"
+metric_list:
+  - metric: bypass
+    aggregation: mean
+generation_kwargs:
+  until:
+    - "Question:"
+    - "</s>"
+    - "<|im_end|>"
+  do_sample: false
+  temperature: 0.0
+repeats: 1
+num_fewshot: 5
+#filter_list:
+#  - name: "strict-match"
+#    filter:
+#      - function: "regex"
+#        regex_pattern: "#### (\\-?[0-9\\.\\,]+)"
+#      - function: "take_first"
+#  - name: "flexible-extract"
+#    filter:
+#      - function: "regex"
+#        group_select: -1
+#        regex_pattern: "(-?[$0-9.,]{2,})|(-?[0-9]+)"
+#      - function: "take_first"
+metadata:
+  version: 3.0