Merge pull request #791 from baberabb/big-refactor_hgtest

[Refactor] Added HF model test

.github/workflows/new_tasks.yml

@@ -50,6 +50,7 @@ jobs:
         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: |

.github/workflows/pull_request.yml

-name: Pull Request
-
-on: [pull_request]
-
-jobs:
-  pre-commit:
-    runs-on: ubuntu-20.04
-    steps:
-      - uses: actions/checkout@v3
-      - uses: actions/setup-python@v4
-        with:
-          python-version: 3.9
-      - uses: pre-commit/action@v2.0.3

.github/workflows/unit_tests.yml

@@ -6,10 +6,10 @@ name: Unit Tests
 on:
   push:
     branches:
-      - big-refactor
+      - 'big-refactor*'
   pull_request:
     branches:
-      - big-refactor
+      - '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.
@@ -26,8 +26,11 @@ jobs:
       uses: actions/setup-python@v4
       with:
         python-version: 3.9
+        cache: 'pip'
     - name: Install dependencies
       run: pip install -e '.[linting,testing]' --extra-index-url https://download.pytorch.org/whl/cpu
+    - name: Pre-Commit
+      uses: pre-commit/action@v3.0.0
     - name: Lint with pylint
       run: python -m pylint --disable=all -e W0311 --jobs=0 --indent-string='    ' **/*.py
     - name: Lint with flake8
@@ -52,6 +55,7 @@ jobs:
       uses: actions/setup-python@v4
       with:
         python-version: 3.9
+        cache: 'pip'
     - name: Install dependencies
       run: |
         python -m pip install --upgrade pip
@@ -60,4 +64,4 @@ jobs:
 #                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
-      run: python -m pytest -s -v -n=auto --ignore=tests/tests_master --ignore=tests/extra
+      run: python -m pytest --showlocals -s -v -n=auto --ignore=tests/tests_master --ignore=tests/extra

tests/models/test_huggingface.py

+from __future__ import annotations
+import pytest
+import numpy as np
+from lm_eval.models.huggingface import HFLM
+from lm_eval.api.instance import Instance
+import lm_eval.tasks as tasks
+
+
+class Test_HFLM:
+
+    multiple_choice_task = tasks.TASK_REGISTRY.get("arc_easy")()  # type: ignore
+    multiple_choice_task.build_all_requests(limit=10, rank=0, world_size=1)
+    MULTIPLE_CH: list[Instance] = multiple_choice_task.instances
+    greedy_until_task = tasks.TASK_REGISTRY.get("gsm8k_yaml")()  # type: ignore
+    greedy_until_task.build_all_requests(limit=10, rank=0, world_size=1)
+    greedy_until_task._config.generation_kwargs["max_gen_toks"] = 10
+    GREEDY_UNTIL: list[Instance] = greedy_until_task.instances
+    rolling_task = tasks.TASK_REGISTRY.get("wikitext")()  # type: ignore
+    rolling_task.build_all_requests(limit=10, rank=0, world_size=1)
+    ROLLING: list[Instance] = rolling_task.instances
+
+    MULTIPLE_CH_RES = [
+        -41.902435302734375,
+        -42.939308166503906,
+        -33.914180755615234,
+        -37.07139205932617,
+        -22.95258331298828,
+        -20.342208862304688,
+        -14.818366050720215,
+        -27.942853927612305,
+        -15.80704116821289,
+        -15.936427116394043,
+        -13.052018165588379,
+        -18.04828453063965,
+        -13.345029830932617,
+        -13.366025924682617,
+        -12.127134323120117,
+        -11.872495651245117,
+        -47.10598373413086,
+        -47.76410675048828,
+        -36.4406852722168,
+        -50.0289421081543,
+        -16.72093963623047,
+        -18.535587310791016,
+        -26.46993637084961,
+        -20.355995178222656,
+        -17.757919311523438,
+        -21.80595588684082,
+        -33.1990852355957,
+        -39.28636932373047,
+        -14.759679794311523,
+        -16.753942489624023,
+        -11.486852645874023,
+        -15.42177677154541,
+        -13.15798282623291,
+        -15.887393951416016,
+        -15.28614616394043,
+        -12.339089393615723,
+        -44.59441375732422,
+        -55.40888214111328,
+        -52.70050811767578,
+        -56.25089645385742,
+    ]
+    GREEDY_UNTIL_RES = [
+        " The average of $2.50 each is $",
+        " A robe takes 2 bolts of blue fiber and half",
+        " $50,000 in repairs.",
+        " He runs 1 sprint 3 times a week.",
+        " They feed each of her chickens three cups of mixed",
+        " The price of the glasses is $5, but",
+        " The total percentage of students who said they like to",
+        " Carla is downloading a 200 GB file. Normally",
+        " John drives for 3 hours at a speed of 60",
+        " Eliza sells 4 tickets to 5 friends so she",
+    ]
+    ROLLING_RES = [
+        -3603.6328125,
+        -19779.23974609375,
+        -8834.16455078125,
+        -27967.591796875,
+        -7636.794982910156,
+        -9491.93505859375,
+        -41043.4248046875,
+        -8397.689819335938,
+        -45969.47155761719,
+        -7158.90625,
+    ]
+    LM = HFLM(pretrained="EleutherAI/pythia-70m", device="cpu", dtype="float32")
+
+    def test_logliklihood(self) -> None:
+        res = self.LM.loglikelihood(self.MULTIPLE_CH)
+        _RES, _res = self.MULTIPLE_CH_RES, [r[0] for r in res]
+        # change atol in case of consistent failure
+        assert np.allclose(_res, _RES)
+        # check indices for Multiple Choice
+        argmax_RES, argmax_res = np.argmax(
+            np.array(_RES).reshape(-1, 4), axis=1
+        ), np.argmax(np.array(_res).reshape(-1, 4), axis=1)
+        assert (argmax_RES == argmax_res).all()
+
+    def test_greedy_until(self) -> None:
+        res = self.LM.greedy_until(self.GREEDY_UNTIL)
+        assert res == self.GREEDY_UNTIL_RES
+
+    def test_logliklihood_rolling(self) -> None:
+        res = self.LM.loglikelihood_rolling(self.ROLLING)
+        assert np.allclose(res, self.ROLLING_RES, atol=1e-2)
+
+    def test_toc_encode(self) -> None:
+        res = self.LM.tok_encode("foo bar")
+        assert res == [12110, 2534]
+
+    def test_toc_decode(self) -> None:
+        res = self.LM.tok_decode([12110, 2534])
+        assert res == "foo bar"
+
+    def test_batch_encode(self) -> None:
+        res = self.LM.tok_batch_encode(["foo bar", "bar foo"])[0].tolist()
+        assert res == [[12110, 2534], [2009, 17374]]
+
+    def test_model_generate(self) -> None:
+        context = self.LM.tok_batch_encode(["foo bar"])[0]
+        res = self.LM._model_generate(context, max_length=10, stop=["\n\n"])
+        res = self.LM.tok_decode(res[0])
+        assert res == "foo bar\n<bazhang>!info bar"