Fix model unittest CI

Author: huanghengheng

.github/workflows/ce-unittest-gpu.yml

@@ -102,7 +102,7 @@ jobs:
           pip config set global.cache-dir "/home/.cache/pip"
           set -e
           cd /workspace/PaddleFormers && git config --global --add safe.directory $PWD
-          timeout 40m bash scripts/unit_test/ci_unit.sh ${paddle_whl} true
+          timeout 40m bash scripts/unit_test/ci_unittest.sh ${paddle_whl} true
           '
           
       - name: Upload Allure-reports & Logs

.github/workflows/debug-unittest-gpu.yml

@@ -96,6 +96,6 @@ jobs:
           echo "work_path: $work_dir/PaddleFormers"
           echo "work_path in docker: /workspace/PaddleFormers"
           echo "cmd: "
-          echo "bash scripts/unit_test/ci_unit.sh ${paddle_whl}"
+          echo "bash scripts/unit_test/ci_unittest.sh ${paddle_whl}"
           echo "or python -m pytest fail_case_name"
           echo "docker rm -f $container_name"

.github/workflows/model-unittest-gpu.yml

@@ -2,7 +2,6 @@ name: Model Unittest GPU CI
 
 on:
   pull_request:
-  push:
   schedule:
     - cron: "0 18 * * *"
   workflow_call:
@@ -112,7 +111,8 @@ jobs:
           cd /workspace/PaddleFormers && git config --global --add safe.directory $PWD
           source $work_dir/../../../proxy
           source $work_dir/../../../AISTUDIO_ACCESS_TOKEN
-          timeout 30m bash scripts/unit_test/ci_model_unit.sh ${paddle_whl}
+          cp -r $work_dir/../../../models ./models
+          timeout 30m bash scripts/unit_test/ci_model_unittest.sh ${paddle_whl}
           '
           
       - name: Upload Products

.github/workflows/unittest-gpu.yml

@@ -2,7 +2,6 @@ name: Unittest GPU CI
 
 on:
   pull_request:
-  push:
   schedule:
     - cron: "0 18 * * *"
   workflow_call:
@@ -56,6 +55,7 @@ jobs:
         run: |
           container_name=${TASK}-$(date +%Y%m%d-%H%M%S)
           echo "container_name=${container_name}" >> "$GITHUB_ENV"
+          echo "Workspace path: ${{ github.workspace }}"
           docker run -d -t --name ${container_name} --net=host -v /dev/shm:/dev/shm --shm-size=32G \
             -v $work_dir/../../..:$work_dir/../../.. \
             -v $work_dir:/workspace \
@@ -112,7 +112,7 @@ jobs:
           cd /workspace/PaddleFormers && git config --global --add safe.directory $PWD
           source $work_dir/../../../proxy
           source $work_dir/../../../AISTUDIO_ACCESS_TOKEN
-          timeout 30m bash scripts/unit_test/ci_unit.sh ${paddle_whl}
+          timeout 30m bash scripts/unit_test/ci_unittest.sh ${paddle_whl}
           '
           
       - name: Upload Products

scripts/regression/test_dpo.py

@@ -26,7 +26,7 @@
 TRAIN_PATH = "./examples"
 CONFIG_PATH = "./examples/config"
 OUTPUT_DIR = tempfile.TemporaryDirectory().name
-MODEL_NAME_OR_PATH = "PaddleFormers/tiny-random-qwen3"
+MODEL_NAME_OR_PATH = "./models/tiny-random-qwen3"
 
 os.environ["NVIDIA_TF32_OVERRIDE"] = "0"
 os.environ["NCCL_ALGO"] = "Tree"
@@ -119,7 +119,7 @@ def test_dpo_full(self):
             train_path,
             updated_config_path,
         ]
-        training_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
+        training_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, timeout=150)
 
         # test training result
         self.dpotrain_tester.assert_result(training_p.returncode, training_p.stdout)
@@ -153,7 +153,7 @@ def test_dpo_lora(self):
             train_path,
             updated_config_path,
         ]
-        training_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
+        training_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, timeout=150)
 
         # test training result
         self.dpotrain_tester.assert_result(training_p.returncode, training_p.stdout)
@@ -177,7 +177,9 @@ def test_dpo_lora(self):
             "--output_path",
             lora_merge_output_dir,
         ]
-        lora_merge_p = subprocess.run(lora_merge_cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
+        lora_merge_p = subprocess.run(
+            lora_merge_cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, timeout=150
+        )
         self.dpotrain_tester.assert_result(lora_merge_p.returncode, lora_merge_p.stdout)
 
         # test lora_merge_model generate

scripts/regression/test_sft.py

@@ -26,7 +26,7 @@
 TRAIN_PATH = "./examples"
 CONFIG_PATH = "./examples/config"
 OUTPUT_DIR = tempfile.TemporaryDirectory().name
-MODEL_NAME_OR_PATH = "PaddleFormers/tiny-random-qwen3"
+MODEL_NAME_OR_PATH = "./models/tiny-random-qwen3"
 
 os.environ["NVIDIA_TF32_OVERRIDE"] = "0"
 os.environ["NCCL_ALGO"] = "Tree"
@@ -123,7 +123,7 @@ def test_sft_full(self):
             train_path,
             updated_config_path,
         ]
-        training_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
+        training_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, timeout=150)
 
         # test training result
         self.sfttrain_tester.assert_result(training_p.returncode, training_p.stdout)
@@ -133,7 +133,7 @@ def test_sft_full(self):
         self.sfttrain_tester.assert_loss(training_p.stdout, EXCEPTED_LOSS)
 
         # test model resume
-        reusme_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
+        reusme_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, timeout=150)
         self.sfttrain_tester.assert_result(reusme_p.returncode, reusme_p.stdout)
 
         EXCEPTED_LOSS = 9.550503
@@ -172,7 +172,7 @@ def test_sft_lora(self):
             train_path,
             updated_config_path,
         ]
-        training_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
+        training_p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, timeout=150)
 
         # test training result
         self.sfttrain_tester.assert_result(training_p.returncode, training_p.stdout)
@@ -200,7 +200,9 @@ def test_sft_lora(self):
             "--output_path",
             lora_merge_output_dir,
         ]
-        lora_merge_p = subprocess.run(lora_merge_cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True)
+        lora_merge_p = subprocess.run(
+            lora_merge_cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, timeout=150
+        )
         self.sfttrain_tester.assert_result(lora_merge_p.returncode, lora_merge_p.stdout)
 
         # test lora_merge_model generate

scripts/unit_test/ci_model_unit.sh renamed to scripts/unit_test/ci_model_unittest.sh

@@ -33,7 +33,7 @@ install_requirements() {
     python -m pip install -r requirements-dev.txt
     python -m pip install -r tests/requirements.txt
     python -m pip uninstall paddlepaddle paddlepaddle_gpu -y
-    python -m pip install --no-cache-dir ${paddle} --no-dependencies
+    python -m pip install --no-cache-dir ${paddle} --no-dependencies --progress-bar off
     python -c "import paddle;print('paddle');print(paddle.__version__);print(paddle.version.show())" >> ${log_path}/commit_info.txt
 
     python setup.py bdist_wheel > /dev/null

scripts/unit_test/ci_unit.sh

@@ -32,7 +32,7 @@ install_requirements() {
     python -m pip install -r requirements-dev.txt
     python -m pip install -r tests/requirements.txt
     python -m pip uninstall paddlepaddle paddlepaddle_gpu -y
-    python -m pip install --no-cache-dir ${paddle} --no-dependencies
+    python -m pip install --no-cache-dir ${paddle} --no-dependencies --progress-bar off
     python -c "import paddle;print('paddle');print(paddle.__version__);print(paddle.version.show())" >> ${log_path}/commit_info.txt
 
     python setup.py bdist_wheel > /dev/null

tests/transformers/auto/test_configuration.py

@@ -26,6 +26,11 @@
 from paddleformers.utils.download import DownloadSource
 from paddleformers.utils.env import CONFIG_NAME
 from tests.testing_utils import set_proxy
+from tests.testing_utils import (
+    require_paddle_at_least_2_gpu,
+    require_paddle_at_least_8_gpu,
+    skip_for_none_ce_case,
+)
 
 from ...utils.test_module.custom_configuration import CustomConfig
 
@@ -79,6 +84,7 @@ def test_from_modelscope(self):
     #     config = AutoConfig.from_pretrained("CompVis/stable-diffusion-v1-4", subfolder="text_encoder")
     #     self.assertEqual(config.hidden_size, 768)
 
+    @skip_for_none_ce_case
     def test_load_from_legacy_config(self):
         number = random.randint(0, 10000)
         legacy_config = {"init_class": "BertModel", "hidden_size": number}
@@ -119,6 +125,7 @@ def test_from_pretrained_cache_dir(self):
             # check against double appending model_name in cache_dir
             self.assertFalse(os.path.exists(os.path.join(tempdir, model_id, model_id)))
 
+    @skip_for_none_ce_case
     def test_load_from_custom_arch(self):
         config_dict = {
             "alibi": False,

tests/transformers/test_configuration_utils.py

@@ -23,6 +23,11 @@
     PretrainedConfig,
     attribute_map,
 )
+from tests.testing_utils import (
+    require_paddle_at_least_2_gpu,
+    require_paddle_at_least_8_gpu,
+    skip_for_none_ce_case,
+)
 from paddleformers.transformers.model_utils import PretrainedModel
 from paddleformers.utils import CONFIG_NAME
 from paddleformers.utils.download import DownloadSource
@@ -154,6 +159,7 @@ def test_from_pretrained_cache_dir(self):
             # check against double appending model_name in cache_dir
             self.assertFalse(os.path.exists(os.path.join(tempdir, model_id, model_id)))
 
+    @skip_for_none_ce_case
     @set_proxy(DownloadSource.HUGGINGFACE)
     def test_load_from_hf(self):
         """test load config from hf"""