Intel ci (#2630)

* Intel CI ?

* Let's try non sharded gemma.

* Snapshot rename

* Apparently container can be gone already.

.github/workflows/build.yaml

@@ -75,10 +75,10 @@ jobs:
                 export label_extension="-intel-cpu"
                 export docker_devices="none"
                 export docker_volume="/mnt/cache"
-                export runs_on="ubuntu-latest"
+                # export runs_on="ubuntu-latest"
-                # export runs_on="aws-highmemory-32-plus-priv"
+                export runs_on="aws-highmemory-32-plus-priv"
                 export platform="cpu"
-                export extra_pytest="-k test_flash_llama_load"
+                export extra_pytest="-k test_flash_gemma_simple"
                 ;;
           esac
           echo $dockerfile

integration-tests/conftest.py

@@ -572,7 +572,10 @@ def launcher(event_loop):
             print(container_output, file=sys.stderr)
 
         finally:
-            container.remove()
+            try:
+                container.remove()
+            except Exception:
+                pass
 
     if DOCKER_IMAGE is not None:
         return docker_launcher

integration-tests/models/test_flash_gemma.py

@@ -16,7 +16,7 @@ async def flash_gemma(flash_gemma_handle):
 @pytest.mark.release
 @pytest.mark.asyncio
 @pytest.mark.private
-async def test_flash_gemma(flash_gemma, response_snapshot):
+async def test_flash_gemma_simple(flash_gemma, response_snapshot):
     response = await flash_gemma.generate(
         "Test request", max_new_tokens=10, decoder_input_details=True
     )

integration-tests/models/test_flash_llama.py

@@ -15,7 +15,7 @@ async def flash_llama(flash_llama_handle):
 
 @pytest.mark.asyncio
 @pytest.mark.private
-async def test_flash_llama(flash_llama, response_snapshot):
+async def test_flash_llama_simple(flash_llama, response_snapshot):
     response = await flash_llama.generate(
         "Test request", max_new_tokens=10, decoder_input_details=True
     )