From b1b2c4115a34ee5d941b37e53c33749f5a789aa2 Mon Sep 17 00:00:00 2001
From: Nan Wang <nan.wang@jina.ai>
Date: Wed, 22 Apr 2020 00:36:28 +0800
Subject: [PATCH 1/2] fix(drivers): add handling for blob

---
 jina/drivers/craft.py                              | 12 +++++++++---
 jina/executors/indexers/vector/numpy.py            |  5 +++--
 tests/executors/encoders/image/test_torchvision.py |  1 +
 3 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/jina/drivers/craft.py b/jina/drivers/craft.py
index 705dce6379dd3..379a42a1c9b64 100644
--- a/jina/drivers/craft.py
+++ b/jina/drivers/craft.py
@@ -2,7 +2,7 @@
 import random
 
 from . import BaseExecutableDriver
-from .helper import array2blob, pb_obj2dict
+from .helper import array2blob, pb_obj2dict, blob2array
 
 
 class BaseCraftDriver(BaseExecutableDriver):
@@ -26,10 +26,16 @@ def __call__(self, *args, **kwargs):
                 continue
             _chunks_to_add = []
             for c in d.chunks:
-                ret = self.exec_fn(**pb_obj2dict(c, self.exec.required_keys))
+                _args_dict = pb_obj2dict(c, self.exec.required_keys)
+                if 'blob' in self.exec.required_keys:
+                    _args_dict['blob'] = blob2array(c.blob)
+                ret = self.exec_fn(**_args_dict)
                 if isinstance(ret, dict):
                     for k, v in ret.items():
-                        setattr(c, k, v)
+                        if k == 'blob':
+                            c.blob.CopyFrom(array2blob(v))
+                        else:
+                            setattr(c, k, v)
                     continue
                 if isinstance(ret, list):
                     for chunk_dict in ret:
diff --git a/jina/executors/indexers/vector/numpy.py b/jina/executors/indexers/vector/numpy.py
index f19c61541821f..5faa4c99a6ba9 100644
--- a/jina/executors/indexers/vector/numpy.py
+++ b/jina/executors/indexers/vector/numpy.py
@@ -132,8 +132,9 @@ def query(self, keys: np.ndarray, top_k: int, *args, **kwargs) -> Tuple['np.ndar
 
         idx = dist.argsort(axis=1)[:, :top_k]
         dist = np.take_along_axis(dist, idx, axis=1)
-
-        return self.int2ext_key[idx], dist
+        _min = dist.min()
+        _max = dist.max()
+        return self.int2ext_key[idx], dist #(_max - dist) / (_max - _min)
 
 
 def _ext_arrs(A, B):
diff --git a/tests/executors/encoders/image/test_torchvision.py b/tests/executors/encoders/image/test_torchvision.py
index 1bcd5478df210..01cbb30731b6d 100644
--- a/tests/executors/encoders/image/test_torchvision.py
+++ b/tests/executors/encoders/image/test_torchvision.py
@@ -7,6 +7,7 @@
 class MyTestCase(ImageTestCase):
     def _get_encoder(self):
         self.target_output_dim = 1280
+        self.input_dim = 224
         return ImageTorchEncoder()
 
 

From fa1a44954cf0a5967989c13e052d705e698ab46b Mon Sep 17 00:00:00 2001
From: Nan Wang <nan.wang@jina.ai>
Date: Wed, 22 Apr 2020 00:40:46 +0800
Subject: [PATCH 2/2] fix: undo the changes

---
 jina/executors/indexers/vector/numpy.py | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/jina/executors/indexers/vector/numpy.py b/jina/executors/indexers/vector/numpy.py
index 5faa4c99a6ba9..fe0db2cfd9a66 100644
--- a/jina/executors/indexers/vector/numpy.py
+++ b/jina/executors/indexers/vector/numpy.py
@@ -132,9 +132,7 @@ def query(self, keys: np.ndarray, top_k: int, *args, **kwargs) -> Tuple['np.ndar
 
         idx = dist.argsort(axis=1)[:, :top_k]
         dist = np.take_along_axis(dist, idx, axis=1)
-        _min = dist.min()
-        _max = dist.max()
-        return self.int2ext_key[idx], dist #(_max - dist) / (_max - _min)
+        return self.int2ext_key[idx], dist
 
 
 def _ext_arrs(A, B):