Add question generator to example

examples/pytorch/question-generator/cortex.yaml

@@ -0,0 +1,10 @@
+# WARNING: you are on the master branch; please refer to examples on the branch corresponding to your `cortex version` (e.g. for version 0.19.*, run `git checkout -b 0.19` or switch to the `0.19` branch on GitHub)
+
+- name: question-generator
+  kind: RealtimeAPI
+  predictor:
+    type: python
+    path: predictor.py
+  compute:
+    cpu: 1
+    mem: 6G

examples/pytorch/question-generator/dependencies.sh

@@ -0,0 +1,4 @@
+# WARNING: you are on the master branch; please refer to examples on the branch corresponding to your `cortex version` (e.g. for version 0.19.*, run `git checkout -b 0.19` or switch to the `0.19` branch on GitHub)
+
+# torchvision isn’t required for this example, and pip was throwing warnings with it installed
+pip uninstall torchvision -y

examples/pytorch/question-generator/predictor.py

@@ -0,0 +1,36 @@
+# WARNING: you are on the master branch; please refer to examples on the branch corresponding to your `cortex version` (e.g. for version 0.19.*, run `git checkout -b 0.19` or switch to the `0.19` branch on GitHub)
+
+from transformers import AutoModelWithLMHead, AutoTokenizer
+import spacy
+import subprocess
+import json
+
+
+class PythonPredictor:
+    def __init__(self, config):
+        subprocess.call("python -m spacy download en_core_web_sm".split(" "))
+        import en_core_web_sm
+
+        self.tokenizer = AutoTokenizer.from_pretrained(
+            "mrm8488/t5-base-finetuned-question-generation-ap"
+        )
+        self.model = AutoModelWithLMHead.from_pretrained(
+            "mrm8488/t5-base-finetuned-question-generation-ap"
+        )
+        self.nlp = en_core_web_sm.load()
+
+    def predict(self, payload):
+        context = payload["context"]
+        answer = payload["answer"]
+        max_length = int(payload.get("max_length", 64))
+
+        input_text = "answer: {}  context: {} </s>".format(answer, context)
+        features = self.tokenizer([input_text], return_tensors="pt")
+
+        output = self.model.generate(
+            input_ids=features["input_ids"],
+            attention_mask=features["attention_mask"],
+            max_length=max_length,
+        )
+
+        return {"result": self.tokenizer.decode(output[0])}

examples/pytorch/question-generator/requirements.txt

@@ -0,0 +1,4 @@
+spacy==2.1.8
+-e git+https://github.com/huggingface/transformers.git#egg=transformers
+--find-links https://download.pytorch.org/whl/torch_stable.html
+torch==1.6.0+cpu

examples/pytorch/question-generator/sample.json

@@ -0,0 +1,4 @@
+{
+    "context": "Sarah works as a software engineer in London",
+    "answer": "London"
+}