Upload folder using huggingface_hub

arc_commonsense.yaml

@@ -0,0 +1,25 @@
+task: arc_commonsense
+dataset_path: CultriX/arc-challenge-train-100
+task: arc_challenge
+dataset_path: allenai/ai2_arc
+dataset_name: ARC-Challenge
+output_type: multiple_choice
+training_split: train
+validation_split: validation
+test_split: test
+doc_to_text: "Question: {{question}}
+Answer:"
+doc_to_target: "{{choices.label.index(answerKey)}}"
+doc_to_choice: "{{choices.text}}"
+should_decontaminate: true
+doc_to_decontamination_query: "Question: {{question}}
+Answer:"
+metric_list:
+  - metric: acc
+    aggregation: mean
+    higher_is_better: true
+  - metric: acc_norm
+    aggregation: mean
+    higher_is_better: true
+metadata:
+  version: 1.0

mathqa.py

@@ -0,0 +1,5 @@
+
+import re
+def doc_to_text(doc) -> str:
+    problem = doc['question']
+    return f"Problem: {problem}\nAnswer:"

mathqa.yaml

@@ -0,0 +1,41 @@
+task: mathqa
+dataset_path: CultriX/mathqa-train-100
+training_split: train
+validation_split: train
+tag:
+  - math_word_problems
+task: mathqa
+dataset_name: main
+output_type: generate_until
+fewshot_split: train
+test_split: test
+doc_to_text: |-
+  Question: {{question}}
+  Answer:
+doc_to_target: "{{answer}}"
+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
+repeats: 1
+num_fewshot: 5
+filter_list:
+  - name: strict-match
+    filter:
+      - function: regex
+        regex_pattern: '#### (\-?[0-9\.\,]+)'
+      - function: take_first
+metadata:
+  version: 3

mmlu-pro.yaml

@@ -0,0 +1,37 @@
+task: mmlu-pro
+dataset_path: CultriX/mmlu-pro-train-100
+output_type: multiple_choice
+training_split: test
+validation_split: test
+test_split: test
+fewshot_split: validation
+fewshot_config:
+  sampler: first_n
+  doc_to_text: !function utils.fewshot_to_text
+  doc_to_target: ""
+output_type: generate_until
+doc_to_text: !function utils.doc_to_text
+doc_to_target: answer
+filter_list:
+  - name: "custom-extract"
+    filter:
+      - function: "regex"
+        regex_pattern: 'answer is \(?([ABCDEFGHIJ])\)?'
+        # regex_pattern: r".*[aA]nswer:\s*([A-J])",
+      - function: "take_first"
+generation_kwargs:
+  until:
+    - "</s>"
+    - "Q:"
+    - "<|im_end|>"
+  do_sample: false
+  temperature: 0.0
+num_fewshot: 5
+metric_list:
+  - metric: exact_match
+    aggregation: mean
+    higher_is_better: true
+    ignore_case: true
+    ignore_punctuation: true
+metadata:
+  version: 1.0

multiwoz_dialogue.yaml

@@ -0,0 +1,15 @@
+task: multiwoz_dialogue
+dataset_path: CultriX/multiwoz-2turn-train-100
+output_type: generate_until
+training_split: train
+validation_split: train
+test_split: train
+doc_to_text: !function preprocess_multiwoz.doc_to_text
+doc_to_choice: []
+doc_to_target: "{{turns[1].utterance}}"
+metric_list:
+  - metric: bleu
+    aggregation: mean
+    higher_is_better: true
+metadata:
+  version: 1.0

preprocess_arc.py

@@ -0,0 +1,7 @@
+
+def doc_to_text(doc) -> str:
+  choices = ""
+  for idx, choice in enumerate(doc["choices"]["text"]):
+    letter = "ABCDE"[idx]
+    choices+= f"{letter}. {choice}\n"
+  return f"Question: {doc['question']}\n{choices}\nAnswer:"

preprocess_mmlu.py

@@ -0,0 +1,7 @@
+
+def doc_to_text(doc) -> str:
+  choices = ""
+  for idx, choice in enumerate(doc["choices"]):
+    letter = "ABCD"[idx]
+    choices += f"{letter}. {choice}\n"
+  return f"Question: {doc['question']}\n{choices}\nAnswer:"

preprocess_multiwoz.py

@@ -0,0 +1,7 @@
+
+def doc_to_text(doc) -> str:
+  history = ""
+  if len(doc['turns']) > 0:
+    history += f"User: {doc['turns'][0]['utterance']}\n"
+  history += "System: "
+  return history

preprocess_triviaqa.py

@@ -0,0 +1,10 @@
+
+def doc_to_text(doc) -> str:
+  question = doc['question']
+  choices = ""
+  for idx, choice in enumerate(doc['answer']['aliases']):
+    if idx >= 5:
+      break
+    letter = "ABCDE"[idx]
+    choices+= f"{letter}. {choice}\n"
+  return f"Question: {question}\n{choices}Answer:"

qnli.py

@@ -0,0 +1,3 @@
+
+def doc_to_text(doc) -> str:
+    return f"Premise: {doc['premise']}\nHypothesis: {doc['hypothesis']}\nRelation:"

qnli.yaml

@@ -0,0 +1,23 @@
+task: qnli
+dataset_path: CultriX/qnli-train-100
+train_spit: train
+validation_split: validation
+test_split: test
+tag:
+  - qnli
+task: qnli
+tag: glue
+dataset_name: qnli-train-100
+output_type: multiple_choice
+training_split: train
+validation_split: validation
+doc_to_text: "{{question}}
+{{sentence}}
+Question: Does this response answer the question?
+Answer:"
+doc_to_target: label
+doc_to_choice: ["yes", "no"]
+metric_list:
+  - metric: acc
+metadata:
+  version: 1.0

triviaqa_qa.yaml

@@ -0,0 +1,33 @@
+task: triviaqa_qa
+dataset_path: CultriX/triviaqa-rc-train-100
+output_type: multiple_choice
+training_split: train
+validation_split: train
+test_split: train
+doc_to_text: "Question: {{question}}?
+Answer:"
+doc_to_target: "{{answer.aliases}}"
+should_decontaminate: true
+doc_to_decontamination_query: question
+generation_kwargs:
+  until:
+    - "
+"
+    - "."
+    - ","
+  do_sample: false
+  temperature: 0.0
+filter_list:
+  - name: remove_whitespace
+    filter:
+      - function: remove_whitespace
+      - function: take_first
+target_delimiter: " "
+metric_list:
+  - metric: exact_match
+    aggregation: mean
+    higher_is_better: true
+    ignore_case: true
+    ignore_punctuation: true
+metadata:
+  version: 3.0