Add Test for Optimum Intel (#2674)

Author: NoushNabi

.github/workflows/test.yml

@@ -30,6 +30,25 @@ jobs:
       - run: helm-run --run-entries simple1:model=simple/model1 --max-eval-instances 10 --suite test
       - run: helm-summarize --suite test
       - run: helm-server --help
+ 
+  install_openvino:  
+    # Tests that the Optimum Intel command works when only installing required dependencies
+    name: Run Optimum Intel with minimal dependencies only
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.8", "3.9", "3.10"]
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: pip
+      - run: python3 -m pip install --upgrade build
+      - run: python3 -m build
+      - run: python3 -m pip install "$(ls dist/crfm_helm-*.whl)[openvino]"
+      - run: helm-run --run-entries boolq:model=hf-internal-testing/tiny-random-MistralForCausalLM --enable-huggingface-models hf-internal-testing/tiny-random-MistralForCausalLM --suite v1 --max-eval-instances 10 --openvino
 
   test:
     name: Run all tests
@@ -59,3 +78,4 @@ jobs:
       - name: Run entire pipeline quickly without any data
         # Checking RunSpecs with openai/davinci should be comprehensive enough
         run: source venv/bin/activate && helm-run --suite test -m 100 --skip-instances --models-to-run openai/davinci --exit-on-error
+        

src/helm/clients/huggingface_client.py

@@ -74,31 +74,24 @@ def __init__(
                 OpenVINO™ Intermediate Representation (IR) format to accelerate end-to-end pipelines on \
                 Intel® architectures using OpenVINO™ runtime.
                 """
-                from pathlib import Path
                 from helm.common.optional_dependencies import handle_module_not_found_error
 
                 try:
                     from optimum.intel.openvino import OVModelForCausalLM
                 except ModuleNotFoundError as e:
                     handle_module_not_found_error(e, ["openvino"])
 
-                model_file = Path(pretrained_model_name_or_path) / "openvino_model.xml"
-                if model_file.exists():
-                    export = False
-                else:
-                    export = True
-
                 self.device = "cpu"
                 # Security issue: currently we trust remote code by default.
                 # We retain this temporarily to maintain reverse compatibility.
                 # TODO: Delete if-else and don't set trust_remote_code=True
                 if "trust_remote_code" in kwargs:
                     self.model = OVModelForCausalLM.from_pretrained(
-                        pretrained_model_name_or_path, export=export, **kwargs
+                        pretrained_model_name_or_path, export=True, **kwargs
                     ).to(self.device)
                 else:
                     self.model = OVModelForCausalLM.from_pretrained(
-                        pretrained_model_name_or_path, export=export, trust_remote_code=True, **kwargs
+                        pretrained_model_name_or_path, export=True, trust_remote_code=True, **kwargs
                     ).to(self.device)
             else:
                 # Security issue: currently we trust remote code by default.