Merge pull request #693 from baberabb/big-refactor_fixfin

[Refactor] Fix tests

.github/workflows/new_tasks.yml

@@ -8,10 +8,11 @@ on:
     branches:
       - big-refactor
   workflow_dispatch:
-
+# comment/edit out the above to stop/change the triggers
 jobs:
   changed_files:
     runs-on: ubuntu-latest  # windows-latest || macos-latest
+    timeout-minutes: 120
     name: Scan for changed tasks
     steps:
       - name: checkout
@@ -19,11 +20,15 @@ jobs:
         with:
           fetch-depth: 0  # OR "2" -> To retrieve the preceding commit.
 
-      # Example 1
+      # Uses the tj-actions/changed-files@v37 action to check for changes.
+      # Outputs provided here: https://github.com/tj-actions/changed-files#outputs
+      # The `files_yaml` input optionally takes a yaml string to specify filters,
+      # and prepends the filter name to the standard output names.
       - name: Check task folders
         id: changed-tasks
         uses: tj-actions/changed-files@v37.1.2
         with:
+          # tasks checks the tasks folder and api checks the api folder for changes
           files_yaml: |
             tasks:
               - lm_eval/tasks/**
@@ -31,31 +36,35 @@ jobs:
               - lm_eval/api/**
           write_output_files: true
 
+    # The next step is optional; the files are written to the workspace by default (above).
+    # so it's just for debugging
       - name: Run Tests
         if: steps.changed-tasks.outputs.tasks_any_modified == 'true' || steps.changed-tasks.outputs.api_any_modified == 'true'
         run: |
           echo .github/outputs/tasks_all_changed_and_modified_files.txt >> 'GITHUB_ENV'
           echo "One or more test file(s) has changed."
           echo "List of all the files that have changed: ${{ steps.changed-tasks.outputs.tasks_all_modified_files }}"
+
       - name: Set up Python 3.9
         if: steps.changed-tasks.outputs.tasks_any_modified == 'true' || steps.changed-tasks.outputs.api_any_modified == 'true'
         uses: actions/setup-python@v4
         with:
           python-version: 3.9
-          cache: 'pip'
       - name: Install dependencies
         if: steps.changed-tasks.outputs.tasks_any_modified == 'true' || steps.changed-tasks.outputs.api_any_modified == 'true'
         run: |
             python -m pip install --upgrade pip
             pip install -e '.[testing]' --extra-index-url https://download.pytorch.org/whl/cpu
-            #         Install optional git dependencies
-    #                pip install bleurt@https://github.com/google-research/bleurt/archive/b610120347ef22b494b6d69b4316e303f5932516.zip#egg=bleurt
-    #        if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+    #   Install optional git dependencies
+    #       pip install bleurt@https://github.com/google-research/bleurt/archive/b610120347ef22b494b6d69b4316e303f5932516.zip#egg=bleurt
+    #       if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
       - name: Test with pytest
+        # if new tasks are added, run tests on them
         if: steps.changed-tasks.outputs.tasks_any_modified == 'true'
-        run: python -m pytest tests/test_tasks.py -s -vv -n=auto --new_task
+        run: python -m pytest tests/extra/test_new_tasks.py -s -vv -n=auto
+        # if api is modified, run tests on it
       - name: Test more tasks with pytest
         env:
           API: true
         if: steps.changed-tasks.outputs.api_any_modified == 'true'
-        run: python -m pytest tests/test_api.py -s -vv -n=auto --new_task
+        run: python -m pytest tests/extra/test_new_tasks.py -s -vv -n=auto

.github/workflows/unit_tests.yml

 # This workflow will install Python dependencies, run tests and lint with a variety of Python versions
 # For more information see: https://docs.github.com/en/actions/automating-builds-and-tests/building-and-testing-python
-
+# just comment out unwanted steps to turn off the test.
 name: Unit Tests
 
 on:
@@ -11,7 +11,8 @@ on:
     branches:
       - big-refactor
   workflow_dispatch:
-
+# Jobs run concurrently and steps run sequentially within a job.
+# jobs: linter and cpu_tests. Add more jobs/steps as required.
 jobs:
   linter:
     name: Linters
@@ -35,9 +36,10 @@ jobs:
         flake8 . --count --select=F,E9,E71,E72,E501,E112,E113,W6 --extend-ignore=F541 --show-source --statistics --exit-zero
         # exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
         flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
-    - name: Lint with mypy
-      run: mypy . --ignore-missing-imports --check-untyped-defs --explicit-package-bases --warn-unreachable
-
+      # mypy turned off for now
+#    - name: Lint with mypy
+#      run: mypy . --ignore-missing-imports --check-untyped-defs --explicit-package-bases --warn-unreachable
+# Job 2
   testcpu:
     name: CPU Tests
     runs-on: ubuntu-latest

.gitignore

@@ -3,6 +3,15 @@ env
 data/
 lm_cache
 .idea
-
-*.egg-info/
+build
+dist
+*.egg-info
+venv
 .vscode/
+temp
+__pycache__
+.ipynb_checkpoints
+temp
+# IPython
+profile_default/
+ipython_config.py

tests/conftest.py

-def pytest_addoption(parser):
-    parser.addoption(
-        "--new_task",
-        action="store_true",
-        help="new_tasks_found",
-    )

tests/extra/test_new_tasks.py

+import pytest
+from itertools import islice
+import lm_eval.tasks as tasks
+from .utilities_testing import load_changed_files, parser
+from typing import List
+from lm_eval.api.task import ConfigurableTask
+import os
+
+
+# GitHub CI
+def new_tasks() -> List[str]:
+    FILENAME = ".github/outputs/tasks_all_changed_and_modified_files.txt"
+    if os.path.exists(FILENAME):
+        # If tasks folder has changed then we get the list of files from FILENAME
+        # and parse the yaml files to get the task names.
+        return parser(load_changed_files(FILENAME))
+    elif os.getenv("API") is not None:
+        # Or if API has changed then we set the ENV variable API to True
+        # and run  given tasks.
+        return ["arc_easy", "hellaswag", "piqa", "wikitext"]
+    # if both not true just do arc_easy
+    else:
+        return ["arc_easy"]
+
+
+def get_task_class() -> List[ConfigurableTask]:
+    task_name = new_tasks()
+    x = [cls for name, cls in tasks.TASK_REGISTRY.items() if name in task_name]
+    return x
+
+
+@pytest.fixture()
+def limit() -> int:
+    return 10
+
+
+# Tests
+@pytest.mark.parametrize("task_class", get_task_class())
+class TestNewTasks:
+    def test_download(self, task_class: ConfigurableTask):
+        task_class().download()
+        assert task_class().dataset is not None
+
+    def test_has_training_docs(self, task_class: ConfigurableTask):
+        assert task_class().has_training_docs() in [True, False]
+
+    def test_check_training_docs(self, task_class: ConfigurableTask):
+        task = task_class()
+        if task.has_training_docs():
+            assert task._config["training_split"] is not None
+
+    def test_has_validation_docs(self, task_class):
+        assert task_class().has_validation_docs() in [True, False]
+
+    def test_check_validation_docs(self, task_class):
+        task = task_class()
+        if task.has_validation_docs():
+            assert task._config["validation_split"] is not None
+
+    def test_has_test_docs(self, task_class):
+        assert task_class().has_test_docs() in [True, False]
+
+    def test_check_test_docs(self, task_class):
+        task = task_class()
+        if task.has_test_docs():
+            assert task._config["test_split"] is not None
+
+    def test_should_decontaminate(self, task_class):
+        task = task_class()
+        assert task.should_decontaminate() in [True, False]
+        if task.should_decontaminate():
+            assert task._config["doc_to_decontamination_query"] is not None
+
+    def test_doc_to_text(self, task_class, limit):
+        task = task_class()
+        arr = (
+            list(islice(task.test_docs(), limit))
+            if task.has_test_docs()
+            else list(islice(task.validation_docs(), limit))
+        )
+        _array = [task.doc_to_text(doc) for doc in arr]
+        # space convention; allow txt to have length 0 for perplexity-like tasks since the model tacks an <|endoftext|> on
+        assert all(
+            isinstance(x, str) and (x[-1] != " " if len(x) != 0 else True)
+            for x in _array
+        )
+
+    def test_create_choices(self, task_class, limit):
+        task = task_class()
+        arr = (
+            list(islice(task.test_docs(), limit))
+            if task.has_test_docs()
+            else list(islice(task.validation_docs(), limit))
+        )
+        if "multiple_choice" in task._config.group:
+            _array = [task.doc_to_choice(doc) for doc in arr]
+            # assert all(len(x) == 4 for x in _array)
+            assert all(isinstance(x, list) for x in _array)
+            assert all(isinstance(x[0], str) for x in _array)
+
+    def test_doc_to_target(self, task_class, limit):
+        task = task_class()
+        arr = (
+            list(islice(task.test_docs(), limit))
+            if task.has_test_docs()
+            else list(islice(task.validation_docs(), limit))
+        )
+        _array_target = [task.doc_to_target(doc) for doc in arr]
+        assert all(isinstance(label, int) for label in _array_target)
+        assert len(_array_target) == limit if limit else True
+        # _array_text = [task.doc_to_text(doc) for doc in arr]
+        # Not working
+        # assert all(tgt[0] == " " or txt[-1] == "\n" if  len(txt) != 0 else True for txt, tgt in zip(_array_text, _array_target))
+
+    def test_build_all_requests(self, task_class, limit):
+        task_class().build_all_requests(rank=1, limit=limit, world_size=1)
+        assert task_class.instances is not None
+
+    def test_construct_requests(self, task_class, limit):
+        task = task_class()
+        arr = (
+            list(islice(task.test_docs(), limit))
+            if task.has_test_docs()
+            else list(islice(task.validation_docs(), limit))
+        )
+        requests = [task.construct_requests(doc, task.doc_to_text(doc)) for doc in arr]
+        assert all(isinstance(doc, list) for doc in requests)
+        assert len(requests) == limit if limit else True

tests/extra/test_utils.py → tests/extra/utilities_testing.py

@@ -2,16 +2,25 @@ import json
 from typing import List
 from lm_eval.utils import load_yaml_config
 from pathlib import Path
+import sys
 
+# This is the path where the output for the changed files for the tasks folder is stored
+# FILE_PATH = file_path = ".github/outputs/tasks_all_changed_and_modified_files.txt"
 
-FILE_PATH = file_path = ".github/outputs/tasks_all_changed_and_modified_files.txt"
 
-
-def load_changed_files(file_path: str = FILE_PATH) -> List[str]:
+# reads a text file and returns a list of words
+# used to read the output of the changed txt from tj-actions/changed-files
+def load_changed_files(file_path: str) -> List[str]:
     with open(file_path, "r") as f:
-        return [line.strip() for line in f.readlines()]
+        content = f.read()
+        words_list = [x for x in content.split()]
+        sys.stdout.write(f"list of files: {words_list}")
+    return words_list
 
 
+# checks the txt file for list of changed files.
+# if file ends with .yaml then check yaml for task name
+# if file ends with .py then parse the folder for all yaml files
 def parser(full_path: List[str]) -> List[str]:
     _output = set()
     for x in full_path:

tests/test_tasks.py

-import pytest
 from itertools import islice
+import pytest
+from typing import List
 import lm_eval.tasks as tasks
-from tests.extra.test_utils import load_changed_files, parser
-from typing import List, ClassVar
-import os
-
+from lm_eval.api.task import ConfigurableTask
 
+# Using fixtures to get the task class and limit
 @pytest.fixture()
-def any_new_tasks(request) -> bool:
-    return request.config.getoption("--new_task")
-
-
-# ["arc_easy] else get list of new tasks
-def new_tasks(any_new_tasks: bool) -> List[str]:
-    FILENAME = ".github/outputs/tasks_all_changed_and_modified_files.txt"
-    if any_new_tasks and os.path.exists(FILENAME):
-        return [parser(load_changed_files(FILENAME))]
-    elif os.getenv("API") is not None:
-        return ["arc_easy", "hellaswag", "piqa", "wikitext"]
-    else:
-        return ["arc_easy"]
-
-
-@pytest.fixture(params=new_tasks(any_new_tasks))
-def task_class(request):
-    task_name = request.param
-    return [cls for name, cls in tasks.TASK_REGISTRY.items() if name in task_name][0]
+def task_class() -> ConfigurableTask:
+    task_name = ["arc_easy"]
+    x = [cls for name, cls in tasks.TASK_REGISTRY.items() if name in task_name]
+    return x[0]
 
 
 @pytest.fixture()
-def limit(any_new_tasks: bool) -> int:
-    return 100 if any_new_tasks else 10
+def limit() -> int:
+    return 10
 
 
 # Tests
 
 
-def test_download(task_class):
+def test_download(task_class: ConfigurableTask):
     task_class().download()
     assert task_class().dataset is not None
 
 
-def test_has_training_docs(task_class):
+def test_has_training_docs(task_class: ConfigurableTask):
     assert task_class().has_training_docs() in [True, False]
 
 
-def test_check_training_docs(task_class):
+def test_check_training_docs(task_class: ConfigurableTask):
     task = task_class()
-    assert task.has_training_docs() if task._config["training_split"] else True
+    if task.has_training_docs():
+        assert task._config["training_split"] is not None
 
 
 def test_has_validation_docs(task_class):
-    assert task_class().has_training_docs() in [True, False]
+    assert task_class().has_validation_docs() in [True, False]
 
 
 def test_check_validation_docs(task_class):
     task = task_class()
-    assert (
-        task_class().has_training_docs() if task._config["validation_split"] else True
-    )
+    if task.has_validation_docs():
+        assert task._config["validation_split"] is not None
 
 
 def test_has_test_docs(task_class):
-    assert task_class().has_training_docs() in [True, False]
+    assert task_class().has_test_docs() in [True, False]
 
 
 def test_check_test_docs(task_class):
     task = task_class()
-    assert task_class().has_training_docs() if task._config["test_split"] else True
+    if task.has_test_docs():
+        assert task._config["test_split"] is not None
 
 
 def test_should_decontaminate(task_class):
-    task_class = task_class()
-    assert task_class.should_decontaminate() in [True, False]
-    if task_class.should_decontaminate():
-        assert task_class._config["doc_to_decontamination_query"] is not None
+    task = task_class()
+    assert task.should_decontaminate() in [True, False]
+    if task.should_decontaminate():
+        assert task._config["doc_to_decontamination_query"] is not None
 
 
 def test_doc_to_text(task_class, limit):
+    task = task_class()
     arr = (
-        list(islice(task_class().test_docs(), limit))
-        if limit
-        else list(task_class().test_docs())
+        list(islice(task.test_docs(), limit))
+        if task.has_test_docs()
+        else list(islice(task.validation_docs(), limit))
     )
-    _array = [task_class().doc_to_text(doc) for doc in arr]
+    _array = [task.doc_to_text(doc) for doc in arr]
     # space convention; allow txt to have length 0 for perplexity-like tasks since the model tacks an <|endoftext|> on
     assert all(
         isinstance(x, str) and (x[-1] != " " if len(x) != 0 else True) for x in _array
@@ -91,24 +77,27 @@ def test_doc_to_text(task_class, limit):
 
 
 def test_create_choices(task_class, limit):
+    task = task_class()
     arr = (
-        list(islice(task_class().test_docs(), limit))
-        if limit
-        else list(task_class().test_docs())
+        list(islice(task.test_docs(), limit))
+        if task.has_test_docs()
+        else list(islice(task.validation_docs(), limit))
     )
-    _array = [task_class().doc_to_choice(doc) for doc in arr]
-    # assert all(len(x) == 4 for x in _array)
-    assert all(isinstance(x, list) for x in _array)
-    assert all(isinstance(x[0], str) for x in _array)
+    if "multiple_choice" in task._config.group:
+        _array = [task.doc_to_choice(doc) for doc in arr]
+        # assert all(len(x) == 4 for x in _array)
+        assert all(isinstance(x, list) for x in _array)
+        assert all(isinstance(x[0], str) for x in _array)
 
 
 def test_doc_to_target(task_class, limit):
+    task = task_class()
     arr = (
-        list(islice(task_class().test_docs(), limit))
-        if limit
-        else list(task_class().test_target())
+        list(islice(task.test_docs(), limit))
+        if task.has_test_docs()
+        else list(islice(task.validation_docs(), limit))
     )
-    _array_target = [task_class().doc_to_target(doc) for doc in arr]
+    _array_target = [task.doc_to_target(doc) for doc in arr]
     assert all(isinstance(label, int) for label in _array_target)
     assert len(_array_target) == limit if limit else True
     # _array_text = [task.doc_to_text(doc) for doc in arr]
@@ -122,15 +111,13 @@ def test_build_all_requests(task_class, limit):
 
 
 def test_construct_requests(task_class, limit):
+    task = task_class()
     arr = (
-        list(islice(task_class().test_docs(), limit))
-        if limit
-        else list(task_class().test_docs())
+        list(islice(task.test_docs(), limit))
+        if task.has_test_docs()
+        else list(islice(task.validation_docs(), limit))
     )
-    requests = [
-        task_class().construct_requests(doc, task_class().doc_to_text(doc))
-        for doc in arr
-    ]
+    requests = [task.construct_requests(doc, task.doc_to_text(doc)) for doc in arr]
     assert all(isinstance(doc, list) for doc in requests)
     assert len(requests) == limit if limit else True