Add ONNXRuntime support (#157)

haystack/reader/farm.py

@@ -242,7 +242,7 @@ class FARMReader(BaseReader):
 
         # get answers from QA model
         predictions = self.inferencer.inference_from_dicts(
-            dicts=input_dicts, rest_api_schema=True, multiprocessing_chunksize=1
+            dicts=input_dicts, return_json=True, multiprocessing_chunksize=1
         )
         # assemble answers from all the different documents & format them.
         # For the "no answer" option, we collect all no_ans_gaps and decide how likely
@@ -442,3 +442,23 @@ class FARMReader(BaseReader):
             )
         predictions = self.predict(question, documents, top_k)
         return predictions
+
+    @classmethod
+    def convert_to_onnx(cls, model_name_or_path, opset_version: int = 11, optimize_for: Optional[str] = None):
+        """
+        Convert a PyTorch BERT model to ONNX format and write to ./onnx-export dir. The converted ONNX model
+        can be loaded with in the `FARMReader` using the export path as `model_name_or_path` param.
+
+        Usage:
+            >>> from haystack.reader.farm import FARMReader
+            >>> FARMReader.convert_to_onnx(model_name_or_path="deepset/bert-base-cased-squad2", optimize_for="gpu_tensor_core")
+            >>> FARMReader(model_name_or_path=Path("onnx-export"))
+
+
+        :param opset_version: ONNX opset version
+        :param optimize_for: optimize the exported model for a target device. Available options
+                             are "gpu_tensor_core" (GPUs with tensor core like V100 or T4),
+                             "gpu_without_tensor_core" (most other GPUs), and "cpu".
+        """
+        inferencer = Inferencer.load(model_name_or_path, task_type="question_answering")
+        inferencer.model.convert_to_onnx(output_path=Path("onnx-export"), opset_version=opset_version, optimize_for=optimize_for)

requirements.txt

@@ -1,4 +1,4 @@
-farm==0.4.3
+farm==0.4.4
 --find-links=https://download.pytorch.org/whl/torch_stable.html
 fastapi
 uvicorn