Lab Submission

.gitignore

@@ -0,0 +1 @@
+.env

lab-extractive-question-answering.ipynb

@@ -112,7 +112,7 @@
    },
    "outputs": [],
    "source": [
-    "!pip install -qU datasets pinecone-client sentence-transformers torch"
+    "# !pip install -qU datasets pinecone-client sentence-transformers torch"
    ]
   },
   {
@@ -242,7 +242,8 @@
     "from datasets import load_dataset\n",
     "\n",
     "# load the squad dataset into a pandas dataframe\n",
-    "df = load_dataset(\"squad\", split=\"train\").to_pandas()"
+    "df = load_dataset(\"squad\", split=\"train\").to_pandas()\n",
+    "df.head()"
    ]
   },
   {
@@ -260,9 +261,9 @@
    "outputs": [],
    "source": [
     "# select only title and context column\n",
-    "df = None\n",
+    "df = df[[\"title\", \"context\"]]\n",
     "# drop rows containing duplicate context passages\n",
-    "df = None\n",
+    "df = df.drop_duplicates(subset=[\"context\"])\n",
     "df"
    ]
   },
@@ -293,7 +294,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!pip install -qU langchain-pinecone pinecone-notebooks"
+    "# !pip install -qU langchain-pinecone pinecone-notebooks"
    ]
   },
   {
@@ -310,11 +311,9 @@
     "spec = ServerlessSpec(\n",
     "    cloud=\"aws\", region=\"us-east-1\"\n",
     ")\n",
-    "\n",
     "# connect to pinecone environment\n",
     "pc = Pinecone(\n",
-    "    api_key = PINECONE_API_KEY,\n",
-    "    environment='us-east-1'  # find next to API key in console\n",
+    "    api_key = PINECONE_API_KEY\n",
     ")"
    ]
   },
@@ -337,14 +336,18 @@
    },
    "outputs": [],
    "source": [
-    "index_name = None\n",
+    "index_name = \"question-answering\"\n",
     "\n",
     "# check if the extractive-question-answering index exists\n",
-    "if index_name not in pinecone.list_indexes().names():\n",
-    "    # create the index if it does not exist\n",
-    "    None\n",
+    "if not pc.has_index(index_name):\n",
+    "    pc.create_index(\n",
+    "  name=index_name,\n",
+    "  dimension=384,\n",
+    "  metric=\"cosine\",\n",
+    "  spec=spec\n",
+    "    )\n",
     "# connect to extractive-question-answering index we created\n",
-    "index = pinecone.Index(index_name)"
+    "index = pc.Index(index_name)"
    ]
   },
   {
@@ -550,7 +553,7 @@
     "# set device to GPU if available\n",
     "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
     "# load the retriever model from huggingface model hub\n",
-    "retriever = None #use the 'multi-qa-MiniLM-L6-cos-v1' model from HuggingFace to build the retriever\n",
+    "retriever = SentenceTransformer('sentence-transformers/multi-qa-MiniLM-L6-cos-v1')\n",
     "retriever"
    ]
   },
@@ -609,17 +612,17 @@
     "\n",
     "for i in tqdm(range(0, len(df), batch_size)):\n",
     "    # find end of batch\n",
-    "    None\n",
+    "    i_end = min(i + batch_size, len(df))\n",
     "    # extract batch\n",
-    "    None\n",
+    "    batch = df.iloc[i:i_end]\n",
     "    # generate embeddings for batch\n",
-    "    emb = None\n",
+    "    emb = retriever.encode(batch[\"context\"].tolist()).tolist()\n",
     "    # get metadata\n",
-    "    meta = None\n",
+    "    meta = batch.to_dict(orient=\"records\")\n",
     "    # create unique IDs\n",
-    "    ids = None\n",
+    "    ids = [str(idx) for idx in batch.index]\n",
     "    # add all to upsert list\n",
-    "    to_upsert = None\n",
+    "    to_upsert = list(zip(ids, emb, meta))\n",
     "    # upsert/insert these records to pinecone\n",
     "    _ = index.upsert(vectors=to_upsert)\n",
     "\n",
@@ -718,12 +721,24 @@
    },
    "outputs": [],
    "source": [
-    "from transformers import pipeline\n",
+    "from transformers import AutoTokenizer, AutoModelForQuestionAnswering\n",
     "\n",
     "model_name = 'deepset/electra-base-squad2'\n",
-    "# load the reader model into a question-answering pipeline\n",
-    "reader = pipeline(tokenizer=model_name, model=model_name, task='question-answering', device=device)\n",
-    "reader"
+    "\n",
+    "tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
+    "model = AutoModelForQuestionAnswering.from_pretrained(model_name).to(device)\n",
+    "def reader(question, context):\n",
+    "    inputs = tokenizer(question, context, return_tensors='pt', truncation=True, max_length=512).to(device)\n",
+    "    with torch.no_grad():\n",
+    "        outputs = model(**inputs)\n",
+    "    \n",
+    "    start = outputs.start_logits.argmax()\n",
+    "    end = outputs.end_logits.argmax() + 1\n",
+    "    \n",
+    "    answer = tokenizer.convert_tokens_to_string(\n",
+    "        tokenizer.convert_ids_to_tokens(inputs['input_ids'][0][start:end])\n",
+    "    )\n",
+    "    return {\"answer\": answer, \"score\": outputs.start_logits.max().item()}"
    ]
   },
   {
@@ -748,11 +763,11 @@
     "# gets context passages from the pinecone index\n",
     "def get_context(question, top_k):\n",
     "    # generate embeddings for the question\n",
-    "    xq = None\n",
+    "    xq = retriever.encode(question)\n",
     "    # search pinecone index for context passage with the answer\n",
-    "    xc = None\n",
+    "    xc = index.query(vector=xq.tolist(), top_k=5, include_metadata=True)\n",
     "    # extract the context passage from pinecone search result\n",
-    "    c = None\n",
+    "    c = [x['metadata']['context'] for x in xc['matches']]\n",
     "    return c"
    ]
   },
@@ -942,7 +957,7 @@
   },
   "gpuClass": "standard",
   "kernelspec": {
-   "display_name": "Python 3 (ipykernel)",
+   "display_name": ".venv (3.14.0)",
    "language": "python",
    "name": "python3"
   },
@@ -956,7 +971,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.11.8"
+   "version": "3.14.0"
   },
   "papermill": {
    "default_parameters": {},
@@ -970,11 +985,6 @@
    "start_time": "2021-04-15T21:06:38.122812",
    "version": "2.3.3"
   },
-  "vscode": {
-   "interpreter": {
-    "hash": "5fe10bf018ef3e697f9035d60bf60847932a12bface18908407fd371fe880db9"
-   }
-  },
   "widgets": {
    "application/vnd.jupyter.widget-state+json": {
     "00b344135da443ac90e6e6a0f53cfe0f": {