Merge pull request #57 from deepset-ai/new-device-mgmt

update cookbooks for new device mgmt in Sentence Transformers embedders

notebooks/improve-retrieval-by-embedding-metadata.ipynb

@@ -127,7 +127,8 @@
  "from haystack.components.embedders import SentenceTransformersTextEmbedder, SentenceTransformersDocumentEmbedder\n",
  "from haystack.components.writers import DocumentWriter\n",
  "from haystack.document_stores.types import DuplicatePolicy\n",
- "from haystack.components.retrievers.in_memory import InMemoryEmbeddingRetriever"
+ "from haystack.components.retrievers.in_memory import InMemoryEmbeddingRetriever\n",
+ "from haystack.utils import ComponentDevice"
  ]
  },
  {
@@ -145,7 +146,9 @@
  " indexing.add_component(\"splitter\", DocumentSplitter(split_by='sentence', split_length=2))\n",
  "\n",
  " # in the following componente, we can specify the parameter `metadata_fields_to_embed`, with the metadata to embed\n",
- " indexing.add_component(\"doc_embedder\", SentenceTransformersDocumentEmbedder(model=\"thenlper/gte-large\", device=\"cuda:0\", meta_fields_to_embed=metadata_fields_to_embed)\n",
+ " indexing.add_component(\"doc_embedder\", SentenceTransformersDocumentEmbedder(model=\"thenlper/gte-large\",\n",
+ " device=ComponentDevice.from_str(\"cuda:0\"),\n",
+ " meta_fields_to_embed=metadata_fields_to_embed)\n",
  " )\n",
  " indexing.add_component(\"writer\", DocumentWriter(document_store=document_store, policy=DuplicatePolicy.OVERWRITE))\n",
  "\n",
@@ -167,7 +170,8 @@
  "def create_retrieval_pipeline(document_store):\n",
  "\n",
  " retrieval = Pipeline()\n",
- " retrieval.add_component(\"text_embedder\", SentenceTransformersTextEmbedder(model=\"thenlper/gte-large\", device=\"cuda:0\"))\n",
+ " retrieval.add_component(\"text_embedder\", SentenceTransformersTextEmbedder(model=\"thenlper/gte-large\",\n",
+ " device=ComponentDevice.from_str(\"cuda:0\")))\n",
  " retrieval.add_component(\"retriever\", InMemoryEmbeddingRetriever(document_store=document_store, scale_score=False, top_k=3))\n",
  "\n",
  " retrieval.connect(\"text_embedder\", \"retriever\")\n",

notebooks/multilingual_rag_podcast.ipynb

@@ -109,8 +109,9 @@
  "outputs": [],
  "source": [
  "# from haystack.components.audio import LocalWhisperTranscriber\n",
+ "# from haystack.utils import ComponentDevice\n",
  "\n",
- "# whisper = LocalWhisperTranscriber(model=\"small\", device=\"cuda:0\")\n",
+ "# whisper = LocalWhisperTranscriber(model=\"small\", device=ComponentDevice.from_str(\"cuda:0\"),)\n",
  "# whisper.warm_up()\n",
  "# transcription = whisper.run(audio_files=[\"podcast.mp3\"])\n",
  "\n",
@@ -188,6 +189,7 @@
  "from haystack.components.converters import TextFileToDocument\n",
  "from haystack.components.writers import DocumentWriter\n",
  "from haystack.components.preprocessors import DocumentSplitter\n",
+ "from haystack.utils import ComponentDevice\n",
  "\n",
  "# initialize the Document store\n",
  "document_store = QdrantDocumentStore(\n",
@@ -203,7 +205,7 @@
  " \"embedder\",\n",
  " SentenceTransformersDocumentEmbedder(\n",
  " model=\"intfloat/multilingual-e5-large\", # good multilingual model: https://huggingface.co/intfloat/multilingual-e5-large\n",
- " device=\"cuda:0\", # load the model on GPU\n",
+ " device=ComponentDevice.from_str(\"cuda:0\"), # load the model on GPU\n",
  " prefix=\"passage:\", # as explained in the model card (https://huggingface.co/intfloat/multilingual-e5-large#faq), documents should be prefixed with \"passage:\"\n",
  " ))\n",
  "indexing_pipeline.add_component(\"writer\", DocumentWriter(document_store=document_store))\n",
@@ -1035,7 +1037,7 @@
  " \"text_embedder\",\n",
  " SentenceTransformersTextEmbedder(\n",
  " model=\"intfloat/multilingual-e5-large\", # good multilingual model: https://huggingface.co/intfloat/multilingual-e5-large\n",
- " device=\"cuda:0\", # load the model on GPU\n",
+ " device=ComponentDevice.from_str(\"cuda:0\"), # load the model on GPU\n",
  " prefix=\"query:\", # as explained in the model card (https://huggingface.co/intfloat/multilingual-e5-large#faq), queries should be prefixed with \"query:\"\n",
  " ))\n",
  "query_pipeline.add_component(\"retriever\", QdrantEmbeddingRetriever(document_store=document_store))\n",

notebooks/zephyr-7b-beta-for-rag.ipynb

@@ -153,7 +153,8 @@
  "from haystack.components.preprocessors import DocumentCleaner, DocumentSplitter\n",
  "from haystack.components.embedders import SentenceTransformersTextEmbedder, SentenceTransformersDocumentEmbedder\n",
  "from haystack.components.writers import DocumentWriter\n",
- "from haystack.document_stores.types import DuplicatePolicy"
+ "from haystack.document_stores.types import DuplicatePolicy\n",
+ "from haystack.utils import ComponentDevice"
  ]
  },
  {
@@ -207,7 +208,9 @@
  "indexing = Pipeline()\n",
  "indexing.add_component(\"cleaner\", DocumentCleaner())\n",
  "indexing.add_component(\"splitter\", DocumentSplitter(split_by='sentence', split_length=2))\n",
- "indexing.add_component(\"doc_embedder\", SentenceTransformersDocumentEmbedder(model=\"thenlper/gte-large\", device=\"cuda:0\", metadata_fields_to_embed=[\"title\"]))\n",
+ "indexing.add_component(\"doc_embedder\", SentenceTransformersDocumentEmbedder(model=\"thenlper/gte-large\",\n",
+ " device=ComponentDevice.from_str(\"cuda:0\"), \n",
+ " meta_fields_to_embed=[\"title\"]))\n",
  "indexing.add_component(\"writer\", DocumentWriter(document_store=document_store, policy=DuplicatePolicy.OVERWRITE))\n",
  "\n",
  "indexing.connect(\"cleaner\", \"splitter\")\n",
@@ -1541,7 +1544,8 @@
  "outputs": [],
  "source": [
  "rag = Pipeline()\n",
- "rag.add_component(\"text_embedder\", SentenceTransformersTextEmbedder(model=\"thenlper/gte-large\", device=\"cuda:0\"))\n",
+ "rag.add_component(\"text_embedder\", SentenceTransformersTextEmbedder(model=\"thenlper/gte-large\", \n",
+ " device=ComponentDevice.from_str(\"cuda:0\"))\n",
  "rag.add_component(\"retriever\", InMemoryEmbeddingRetriever(document_store=document_store, top_k=5))\n",
  "rag.add_component(\"prompt_builder\", prompt_builder)\n",
  "rag.add_component(\"llm\", generator)\n",