Serve model via REST API

pyproject.toml

@@ -9,6 +9,8 @@ license = "GPL-3.0-or-later"
 python = "3.8.*"
 tensorflow = "^2.4.1"
 biopython = "^1.78"
+fastapi = "^0.66.0"
+uvicorn = "^0.14.0"
 
 [tool.poetry.dev-dependencies]
 isort = "^5.8.0"

src/api.py

@@ -0,0 +1,21 @@
+from fastapi import FastAPI
+from pydantic import BaseModel
+from model import infer_sequence
+
+app = FastAPI()
+
+
+class Input(BaseModel):
+    sequence: str
+
+
+@app.get("/{sequence}")
+async def get_sequence_path(sequence: str):
+    correct_sequence = await infer_sequence(sequence)
+    return {"sequence": correct_sequence}
+
+
+@app.post("/")
+async def get_sequence_body(sequence: Input):
+    correct_sequence = await infer_sequence(sequence.sequence)
+    return {"sequence": correct_sequence}

src/model.py

@@ -51,8 +51,8 @@ def show_metrics(model, eval_dataset, test_dataset) -> None:
     """
     eval_metrics = model.evaluate(eval_dataset, verbose=0)
     test_metrics = model.evaluate(test_dataset, verbose=0)
-    print(f"Final eval metrics - loss: {eval_metrics[0]} - accuracy: {eval_metrics[1]}")
+    print(f"Eval metrics {eval_metrics}")
-    print(f"Final test metrics - loss: {test_metrics[0]} - accuracy: {test_metrics[1]}")
+    print(f"Test metrics {test_metrics}")
 
 
 def train_model(data_file, label_file, seed_value=42) -> None:
@@ -71,7 +71,7 @@ def train_model(data_file, label_file, seed_value=42) -> None:
     model.save("trained_model")
 
 
-def infer_sequence(sequence) -> None:
+async def infer_sequence(sequence) -> None:
     """
     Predict the correct sequence, using the trained model
     """
@@ -82,3 +82,4 @@ def infer_sequence(sequence) -> None:
     encoded_prediction = argmax(prediction, axis=1)
     final_prediction = decode_sequence(encoded_prediction)
     print(f"Error-corrected sequence: {final_prediction}")
+    return final_prediction