Chris Green

Results 2 issues of Chris Green

I have a parquet dataset with a column consisting of serialized tf.Example protobufs. When I write this dataset and read without any compression I have no problems deserializing the protos....

The code suggest it is but when running sync training with gpu there's a failure here because the referenced class does not exist. https://github.com/bytedance/monolith/blob/88095e52234937cded0786f6dbb89f7731f9ec36/monolith/native_training/native_model.py#L877