if numpy array has been write into tfrecord,how do we can decode it by dali

[xiaoyang-coder]

` def _bytes_feature(self, value): return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))

feature[key] = self._bytes_feature(np.array(data, dtype=np.float32).tobytes()) example_proto = tf.train.Example(features=tf.train.Features(feature=feature)) tf_writer.write(example_proto.SerializeToString()) ` Thank you very much.

[JanuszL]

Hi @xiaoyang-coder,

DALI doesn't provide a standalone NumPy decoder. If you want to store data inside TFRecord please do it as a raw memory, not a NumPy array.

[xiaoyang-coder]

@JanuszL Thank you for your reply, will the NumPy decoder be provide in the near future? It may be useful feature and very practical。

[JanuszL]

We don't have such a plan, however, we would be more than happy to review and accept any PR introducing this functionality. It can be based on already available numpy reader and loader. In the meantime, you can check on how DALI can be easily extended by user-provided functionalities.