Failed to train with multiple GPUs in single node #122
Comments
francktcheng
added a commit
that referenced
this issue
Mar 29, 2023
1. patching shared embedding get_shape method by using TensorShape to be compatible with init_from_checkpoint. 2. patching the `get_shape` method of embeddings created by `tf.contrib.feature_column` API in distributed training. (#122)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
User Story
Failed to train with multiple GPUs in single node when we use
from tensorflow.contrib.layers.python.layers import feature_columnDetailed requirements
API Compatibility
Willing to contribute
Yes
The text was updated successfully, but these errors were encountered: