[Bug]: Insert float16 vector to bfloat16 field could be successful

[binbinlv]

Is there an existing issue for this?

• I have searched the existing issues

Describe the bug

Insert float16 vector to bfloat16 field could be successful

Expected Behavior

Report error

Steps/Code To Reproduce behavior

from pymilvus import connections
from pymilvus import DataType
from pymilvus import Partition
from pymilvus import utility
import time
import h5py

import json
from pymilvus import CollectionSchema, FieldSchema, DataType, utility, connections, Collection
collection_name = "test_f16"
connections.connect()

dim = 128
int64_field = FieldSchema(name="int64", dtype=DataType.INT64, is_primary=True)
bfloat16_vector = FieldSchema(name="bfloat16_vector", dtype=DataType.BFLOAT16_VECTOR, dim=dim)


schema = CollectionSchema(fields=[int64_field, bfloat16_vector])
if utility.has_collection(collection_name):
    old_collection = Collection(collection_name)
    old_collection.drop()
    print("drop")

collection = Collection(collection_name, schema=schema)

def gen_bf16_vectors(data):
      bf16_vectors = []
      for i in range(data.shape[0]):
            raw_vector = data[i]
            bf16_vector = tf.cast(raw_vector, dtype=tf.bfloat16).numpy()
            bf16_vectors.append(bf16_vector)
      return bf16_vectors

def gen_fp16_vectors(data):
       fp16_vectors = []
       for i in range(data.shape[0]):
             raw_vector =  data[i]
             fp16_vector = np.array(raw_vector, dtype=np.float16)
             fp16_vectors.append(fp16_vector)
       return fp16_vectors

import numpy as np
import tensorflow as tf

fname = "/home/sift-128-euclidean.hdf5"
f = h5py.File(fname, 'r')
data = f['train']
test = f['test']
neighbors = f['neighbors']
print(data.shape)
print(neighbors.shape)
print(test.shape)

print("begin insert...")
counter = 0
block_num = 100
block_size = int(data.shape[0]/block_num)
start = time.time()

for t in range(block_num):
    print("inserting: ", counter, " to ", counter + block_size )
    #bytes_data = gen_bf16_vectors(data[counter: counter + block_size])
    bytes_data = gen_fp16_vectors(data[counter: counter + block_size])
    #bytes_data = data[counter: counter + block_size]
    entities = [
            [i for i in range(counter, counter + block_size)],
            bytes_data
            ]
    insert_result =  collection.insert(entities)
    print(insert_result)
    counter = counter + block_size
print ("end of insert, cost: ", time.time()-start)

Environment details

- Hardware/Softward conditions (OS, CPU, GPU, Memory):
- Method of installation (Docker, or from source):
- Milvus version (v0.3.1, or v0.4.0):
- Milvus configuration (Settings you made in `server_config.yaml`):

Anything else?

No response

[binbinlv]

/assign @XuanYang-cn

[XuanYang-cn]

/assign @binbinlv
Please help verify 2.5.0rc8

[XuanYang-cn]

/unassign

[XuanYang-cn]

pymilvus-2.4.1rc7