fix bugs

.github/workflows/test_cli_cuda_pytorch.yaml

@@ -47,10 +47,14 @@ jobs:
         run: |
           pip install -e .[testing,diffusers,timm,peft,bitsandbytes,autoawq,auto-gptq]
 
-      - name: Run tests
+      - name: Run cli tests
         run: |
           pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and not (dp or ddp or device_map or deepspeed)"
 
+      - name: Run tests from example folder
+        run: |
+          pytest tests/test_examples.py -x -s -k "cuda and pytorch"
+
   run_cli_cuda_pytorch_multi_gpu_tests:
     if: ${{
       (github.event_name == 'push') ||
@@ -77,14 +81,11 @@ jobs:
         run: |
           pip install -e .[testing,diffusers,timm,peft,deepspeed]
 
-      # - name: Run cli tests (parallel)
-      #   run: |
-      #     pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and (dp or ddp or device_map)"
-
-      # - name: Run cli tests (sequential)
-      #   run: |
-      #     FORCE_SEQUENTIAL=1 pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and (deepspeed_inference)"
+      - name: Run cli tests (parallel)
+        run: |
+          pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and (dp or ddp or device_map)"
 
-      - name: Run test from example folder
+      - name: Run cli tests (sequential)
         run: |
-          pytest tests/test_examples.py -x -s -k "cuda and pytorch"
+          FORCE_SEQUENTIAL=1 pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and (deepspeed_inference)"
+

.github/workflows/test_cli_rocm_pytorch.yaml

@@ -50,9 +50,13 @@ jobs:
         run: |
           pip install -e .[testing,diffusers,timm,peft,autoawq,auto-gptq]
 
-      - name: Run tests
+      # - name: Run cli tests
+      #   run: |
+      #     pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and not (dp or ddp or device_map or deepspeed) and not bnb"
+
+      - name: Run tests from example folder
         run: |
-          pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and not (dp or ddp or device_map or deepspeed) and not bnb"
+          pytest tests/test_examples.py -x -s -k "rocm and pytorch"
 
   run_cli_rocm_pytorch_multi_gpu_tests:
     if: ${{
@@ -84,14 +88,11 @@ jobs:
         run: |
           pip install -e .[testing,diffusers,timm,deepspeed,peft,autoawq,auto-gptq] "deepspeed<0.15"
 
-      # - name: Run cli tests (parallel)
-      #   run: |
-      #     pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and (dp or ddp or device_map)"
-
-      # - name: Run cli tests (sequential)
-      #   run: |
-      #     FORCE_SEQUENTIAL=1 pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and (deepspeed_inference)"
+      - name: Run cli tests (parallel)
+        run: |
+          pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and (dp or ddp or device_map)"
 
-      - name: Run tests from example folder
+      - name: Run cli tests (sequential)
         run: |
-          pytest tests/test_examples.py -x -s -k "cuda and pytorch"
+          FORCE_SEQUENTIAL=1 pytest tests/test_cli.py -x -s -k "cli and cuda and pytorch and (deepspeed_inference)"
+

examples/tei_bge.yaml

@@ -7,10 +7,6 @@ defaults:
 
 name: tei_bert
 
-launcher:
-  device_isolation: true
-  device_isolation_action: warn
-
 backend:
   device: cpu
   model: BAAI/bge-base-en-v1.5

tests/test_examples.py

@@ -207,3 +207,13 @@ def test_cuda_vllm_configs(config_name):
 @pytest.mark.parametrize("script_name", CUDA_PYTORCH_SCRIPTS)
 def test_cuda_pytorch_scripts(script_name):
     execute_python_script(script_name)
+
+ROCM_PYTORCH_CONFIGS = [config for config in CUDA_PYTORCH_CONFIGS if config != "energy_star.yaml"]
+
+@pytest.mark.parametrize("config_name", ROCM_PYTORCH_CONFIGS)
+def test_rocm_pytorch_configs(config_name):
+    test_yaml_config(config_name)
+
+@pytest.mark.parametrize("script_name", CUDA_PYTORCH_SCRIPTS)
+def test_rocm_pytorch_scripts(script_name):
+    execute_python_script(script_name)