[GHA] Replaced visual_language_chat_sample-ubuntu-llava

.github/workflows/causal_lm_cpp.yml

@@ -399,18 +399,6 @@ jobs:
             f.write(content.encode("utf-8"))
       - run: diff cpp2.txt py2.txt
 
-  visual_language_chat_sample-ubuntu-llava_1_5:
-    uses: ./.github/workflows/job_vlm_sample_llava.yml
-    with:
-      model_id: llava-hf/llava-1.5-7b-hf
-      model_dir: llava_1_5_7b_ov
-
-  visual_language_chat_sample-ubuntu-llava_next:
-    uses: ./.github/workflows/job_vlm_sample_llava.yml
-    with:
-      model_id: llava-hf/llava-v1.6-mistral-7b-hf
-      model_dir: llava_v1_6_mistral_7b_ov
-
   visual_language_chat_sample-ubuntu-internvl2:
     runs-on: ubuntu-22.04-16-cores
     steps:

.github/workflows/linux.yml

@@ -24,6 +24,7 @@ env:
   SCCACHE_SERVER_PORT: 35555
   SCCACHE_CACHE_SIZE: 30G
   SCCACHE_AZURE_KEY_PREFIX: genai/ubuntu/22_04/x64
+  HF_HOME: /mount/caches/huggingface
   GENAI_ARCHIVE_NAME: genai.tar.gz
   GENAI_SAMPLES_NAME: genai_samples.tar.gz
 
@@ -326,16 +327,22 @@ jobs:
           - name: 'LLM'
             marker: 'llm'
             cmd: 'tests/python_tests/samples'
+            runner: 'aks-linux-4-cores-16gb'
           - name: 'Whisper'
             marker: 'whisper'
             cmd: 'tests/python_tests/samples'
+            runner: 'aks-linux-4-cores-16gb'
+          - name: 'VLM'
+            marker: 'vlm'
+            cmd: 'tests/python_tests/samples'
+            runner: 'aks-linux-8-cores-64gb'
 
     needs: [ openvino_download, genai_build_cmake, genai_build_wheel, genai_build_samples ] 
     timeout-minutes: 45
     defaults:
       run:
         shell: bash
-    runs-on: aks-linux-4-cores-16gb
+    runs-on: ${{ matrix.test.runner }}
     container:
       image: openvinogithubactions.azurecr.io/ov_test/ubuntu_22_04_x64:${{ needs.openvino_download.outputs.docker_tag }}
       volumes:

tests/python_tests/pytest.ini

@@ -8,11 +8,13 @@ markers =
     ; samples     - Tests related to the sample models.
     ; llm         - Tests related to large language models.
     ; whisper     - Tests related to the Whisper model.
+    ; vlm         - Tests related to the VLM model.
     precommit
     nightly
     real_models
     samples
     llm
     whisper
+    vlm
 
 addopts = -m precommit

tests/python_tests/samples/conftest.py

@@ -52,12 +52,21 @@
     "TinyStories-1M": {
         "name": "roneneldan/TinyStories-1M",
         "convert_args": ['--trust-remote-code', '--weight-format', 'fp16']
+    },
+    "llava-1.5-7b-hf": {
+        "name": "llava-hf/llava-1.5-7b-hf",
+        "convert_args": ['--trust-remote-code']
+    },    
+    "llava-v1.6-mistral-7b-hf": {
+        "name": "llava-hf/llava-v1.6-mistral-7b-hf",
+        "convert_args": ['--trust-remote-code']
     }
 }
 
 TEST_FILES = {
     "how_are_you_doing_today.wav": "https://storage.openvinotoolkit.org/models_contrib/speech/2021.2/librispeech_s5/how_are_you_doing_today.wav",
-    "adapter_model.safetensors": "https://huggingface.co/smangrul/tinyllama_lora_sql/resolve/main/adapter_model.safetensors"
+    "adapter_model.safetensors": "https://huggingface.co/smangrul/tinyllama_lora_sql/resolve/main/adapter_model.safetensors",
+    "monalisa.jpg": "https://llava-vl.github.io/static/images/monalisa.jpg",
 }
 
 SAMPLES_PY_DIR = os.environ.get("SAMPLES_PY_DIR", os.path.abspath(os.path.join(os.path.dirname(__file__), "../../../samples/python")))

tests/python_tests/samples/test_utils.py

@@ -3,8 +3,10 @@
 from conftest import logger
 import subprocess # nosec B404
 
-def run_sample(command):
+def run_sample(command, input_data=None):
     logger.info(f"Running sample command: {' '.join(command)}")
-    result = subprocess.run(command, capture_output=True, text=True, check=True, encoding='utf-8')
+    if input_data:
+        logger.info(f"Input data: {input_data}")
+    result = subprocess.run(command, capture_output=True, text=True, check=True, encoding='utf-8', input=input_data)
     logger.info(f"Sample output: {result.stdout}")
     return result

tests/python_tests/samples/test_visual_language_chat.py

@@ -0,0 +1,35 @@
+# Copyright (C) 2025 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import os
+import pytest
+import sys
+
+from conftest import TEST_FILES, SAMPLES_PY_DIR, SAMPLES_CPP_DIR
+from test_utils import run_sample
+
+class TestVisualLanguageChat:
+    @pytest.mark.vlm
+    @pytest.mark.samples
+    @pytest.mark.parametrize(
+        "convert_model, sample_args",
+        [
+            pytest.param("llava-1.5-7b-hf", 'Who drew this painting?\nWhen did the painter live?'),
+            pytest.param("llava-v1.6-mistral-7b-hf", 'Who drew this painting?\nWhen did the painter live?'),
+        ],
+        indirect=["convert_model"],
+    )
+    @pytest.mark.parametrize("download_test_content", [TEST_FILES["monalisa.jpg"]], indirect=True)
+    def test_sample_visual_language_chat(self, convert_model, download_test_content, sample_args):
+        # Test Python sample
+        py_script = os.path.join(SAMPLES_PY_DIR, "visual_language_chat/visual_language_chat.py")
+        py_command = [sys.executable, py_script, convert_model, download_test_content]
+        py_result = run_sample(py_command, sample_args)
+
+        # Test CPP sample
+        cpp_sample = os.path.join(SAMPLES_CPP_DIR, 'visual_language_chat')
+        cpp_command =[cpp_sample, convert_model, download_test_content]
+        cpp_result = run_sample(cpp_command, sample_args)
+
+        # Compare results
+        assert py_result.stdout == cpp_result.stdout, f"Results should match"