docs(helper): add docstring for types (#98)

* docs(helper): add docstring for types

* docs(helper): add docstring for types

finetuner/helper.py

@@ -2,22 +2,30 @@
 
 from jina import Document, DocumentArray, DocumentArrayMemmap
 
-AnyDNN = TypeVar('AnyDNN')  #: Any implementation of a Deep Neural Network object
-AnyDataLoader = TypeVar('AnyDataLoader')  #: Any implementation of a data loader
+AnyDNN = TypeVar(
+    'AnyDNN'
+)  #: The type of any implementation of a Deep Neural Network object
+AnyDataLoader = TypeVar(
+    'AnyDataLoader'
+)  #: The type of any implementation of a data loader
 DocumentSequence = TypeVar(
     'DocumentSequence',
     Sequence[Document],
     DocumentArray,
     DocumentArrayMemmap,
     Iterator[Document],
-)
+)  #: The type of any sequence of Document
 DocumentArrayLike = Union[
     DocumentSequence,
     Callable[..., DocumentSequence],
-]
-
-EmbeddingLayerInfoType = List[Dict[str, Any]]
-TunerReturnType = Dict[str, Dict[str, Any]]
+]  #: The type :py:data:`DocumentSequence` or a function that gives :py:data:`DocumentSequence`
+
+EmbeddingLayerInfoType = List[
+    Dict[str, Any]
+]  #: The type of embedding layer information used in Tailor
+TunerReturnType = Dict[
+    str, Dict[str, Any]
+]  #: The type of loss, metric information Tuner returns
 
 
 def get_framework(dnn_model: AnyDNN) -> str:

tests/integration/fit/test_fit_mlp.py

@@ -4,7 +4,7 @@
 import tensorflow as tf
 import torch
 
-import finetuner as jft
+import finetuner
 from finetuner.toydata import generate_fashion_match
 
 
@@ -39,7 +39,7 @@ def test_fit_all(tmpdir):
 
     for kb, b in embed_models.items():
         for h in ['CosineLayer', 'TripletLayer']:
-            result = jft.fit(
+            result = finetuner.fit(
                 b(),
                 head_layer=h,
                 train_data=lambda: generate_fashion_match(

tests/integration/torch/test_torch_trainer.py

@@ -27,8 +27,10 @@ def test_simple_sequential_model(tmpdir, params, head_layer):
 
     # fit and save the checkpoint
     pt.fit(
-        train_data=lambda: fmdg(num_total=params['num_train']),
-        eval_data=lambda: fmdg(num_total=params['num_eval'], is_testset=True),
+        train_data=lambda: fmdg(num_pos=10, num_neg=10, num_total=params['num_train']),
+        eval_data=lambda: fmdg(
+            num_pos=10, num_neg=10, num_total=params['num_eval'], is_testset=True
+        ),
         epochs=params['epochs'],
         batch_size=params['batch_size'],
     )