change the ASR model id

change the ASR model id change the ASR model id change the ASR model id change the ASR model id change the ASR model id change the ASR model id change the ASR model id change the ASR model id
openvinotoolkit · Dec 18, 2024 · 7290856 · 7290856
1 parent dfd4aed
commit 7290856
Showing 1 changed file with 13 additions and 10 deletions.
diff --git a/notebooks/multimodal-rag/multimodal-rag-llamaindex.ipynb b/notebooks/multimodal-rag/multimodal-rag-llamaindex.ipynb
@@ -69,10 +69,12 @@
     }
    ],
    "source": [
+    "%pip uninstall -q -y \"moviepy\" \"decorator\"\n",
+    "\n",
     "%pip install -q \"llama-index-core\" \"llama-index-embeddings-openvino>=0.4.1\" \"llama-index-multi-modal-llms-openvino\" \"llama-index-readers-file\" \\\n",
     "    \"llama-index-vector-stores-qdrant\"  \\\n",
     "    \"transformers>=4.45\" \\\n",
-    "    \"moviepy==1.0.3\" \\\n",
+    "    \"moviepy>=2.1.1\" \\\n",
     "    \"librosa\" \\\n",
     "    \"python-ffmpeg<=1.0.16\" \\\n",
     "    \"open_clip_torch\" \\\n",
@@ -82,7 +84,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": null,
    "id": "f24c553d-8c4c-40dd-86e4-313b99f928d0",
    "metadata": {},
    "outputs": [
@@ -146,7 +148,7 @@
    "source": [
     "import huggingface_hub as hf_hub\n",
     "\n",
-    "asr_model_id = \"OpenVINO/distil-large-v3-int8-ov\"\n",
+    "asr_model_id = \"OpenVINO/distil-whisper-large-v3-int8-ov\"\n",
     "asr_model_path = asr_model_id.split(\"/\")[-1]\n",
     "\n",
     "if not Path(asr_model_path).exists():\n",
@@ -296,17 +298,17 @@
    "outputs": [],
    "source": [
     "video_url = \"https://github.com/user-attachments/assets/2e38afa5-ce3e-448c-ad1a-0e9471039ff1\"\n",
-    "output_video_path = \"./video_data/\"\n",
     "output_folder = \"./mixed_data/\"\n",
     "output_audio_path = \"./mixed_data/output_audio.wav\"\n",
+    "filepath = \"video_data/input_vid.mp4\"\n",
     "\n",
-    "filepath = output_video_path + \"input_vid.mp4\"\n",
+    "example_path = Path(filepath)\n",
+    "example_path.parent.mkdir(parents=True, exist_ok=True)\n",
     "Path(output_folder).mkdir(parents=True, exist_ok=True)\n",
     "\n",
-    "if not Path(filepath).exists():\n",
-    "    r = requests.get(video_url)\n",
-    "    with Path(filepath).open(\"wb\") as f:\n",
-    "        f.write(r.content)"
+    "r = requests.get(video_url)\n",
+    "with example_path.open(\"wb\") as f:\n",
+    "    f.write(r.content)"
    ]
   },
   {
@@ -442,7 +444,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 35,
+   "execution_count": null,
    "id": "8f45641a",
    "metadata": {},
    "outputs": [
@@ -875,6 +877,7 @@
     "    model_id_or_path=vlm_int4_model_path,\n",
     "    device=vlm_device.value,\n",
     "    messages_to_prompt=messages_to_prompt,\n",
+    "    trust_remote_code=True,\n",
     "    generate_kwargs={\"do_sample\": False, \"eos_token_id\": processor.tokenizer.eos_token_id},\n",
     ")"
    ]