Kingsley Liu

@tilakrayal Any progress here?

@gadagashwini Any progress here?

@rchao Any progress here?

@jszaday Thanks for the updates. Please let me know when there is a solution or workaround for this issue.

So the loaded model graph can still be executed correctly under the Mirrored Strategy and the real problem is `tf.print`? Is there any way to verify it is working correctly...

@jszaday I checked the output tensor of the loaded model graph (`print(vector) `after `logit, vector = strategy.run(...)`), and the log is like this: ```bash ------------------------------ step 0 ------------------------------ global_replica_id: Tensor("demo/test_layer/replica_id_in_sync_group:0",...

Hi @ayushjn20, thanks for your feature request. - Workaround The workaround is to perceive `impression_video_id` and `watched_video_ids` as a whole (assume that `impression_video_id` is univalent and `watched_video_ids` is 10-multivalent, and...

@ayushjn20 Glad to hear that the workaround helps. As you have mentioned, we can only get the mean of the embeddings of 10 `watched_video_ids` if we put them in the...

The Parquet data generator issue has been fixed in 22.07 release.

Hi @AtekiRyu , I don't think the Reshape layer is used correctly: ```bash model.add(hugectr.DenseLayer(layer_type = hugectr.Layer_t.Reshape, bottom_names = ["sparse_embedding1"], top_names = ["reshape1"])) #leading_dim=416)) ``` (when leading_dim is commented out, the...