Meet DataTypeNotSupportException when add bert_as_service output as index.

[ws719547997]

Hello everyone,
when I try this:

dim = 768
default_fields = [
    FieldSchema(name="count", dtype=DataType.INT64, is_primary=True),
    FieldSchema(name="float_vector", dtype=DataType.FLOAT_VECTOR, dim=dim)
]
# ...
vec = bc.encode(['今天是个好天气，不冷也不热可是我穿的太厚了'])[0]
collection.insert([id, vec])

then, when collection.insert milvus tells me:

~/anaconda3/envs/torch/lib/python3.6/site-packages/pymilvus/orm/schema.py in parse_fields_from_data(datas)
    279     for d in datas:
    280         if not is_list_like(d):
--> 281             raise DataTypeNotSupportException(0, ExceptionsMessage.DataTypeNotSupport)
    282         d_type = infer_dtype_bydata(d[0])
    283         fields.append(FieldSchema("", d_type))

DataTypeNotSupportException: <DataTypeNotSupportException: (code=0, message=Data type is not support.)>

I found the sourec code, it tells me:

>>> is_list_like(np.array([2]))
        True
>>> is_list_like(np.array(2))
        False

the 'vec' looks like this,

array([ 5.33684731e-01,  2.31654704e-01,  2.40789890e-01,  1.39630705e-01,
         ...
       -8.82823467e-02,  3.08212638e-02,  7.05534071e-02, -9.86864090e-01],
      dtype=float32)

I think it's a np.array, BUT, WHYYYYYYYY ?
明明一切看起来都是没问题的，究竟是问什么呢TAT。

[yhmo]

The insert() requires each field data be a "list of list", even you only insert one row.
Call insert() like this:
collection.insert([ [id], [vec] ])

[ws719547997]

thanks,
Your answer helps me a lot.
code works now!

 collection.insert(
        [
            [38349,38350],
            bc.encode(['今天我穿的薄了，但是今天阴天','研究进展不错但是明天实习就结束啦']).tolist()
        ]
    )
(insert count: 2, delete count: 0, upsert count: 0, timestamp: 428711567337979906)