Deyu Fu issues

Results 5 issues of


                                            Deyu Fu

Allow custom initializer to work through wrapper without requiring @tf.keras.utils.register_keras_serializable()

Current serialize/deserialize code in init is not neccessay

Add api supporting input hash and embedding together

TFDE doesn't complain when it is passed more inputs than would be expected

Currently if more input is supplied, only leading ones will be used and rest will be silently ignored. Need better check and message

[DEV] fix functional test and re-merge branch tree

# What does this PR do ? fix a failed functional test due to main branch api change Also when merging https://github.com/NVIDIA/Megatron-LM/pull/2395, commits are squashed and thus main/dev are no...

Add muon and layerwise distributed optimizer

# What does this PR do ? Add muon and layerwise distributed optimizer to maim ## Contribution process ```mermaid flowchart LR A[Pre-checks] --> B[PR Tests] subgraph Code Review/Approval C1[Expert Review]...