Sandeep Subramanian
Sandeep Subramanian
Signed-off-by: MaximumEntropy # What does this PR do ? Adds support for Enc-Dec models with RPE and PP > 2 **Collection**: NLP # Changelog - Add specific line by line...
# What does this PR do ? Adds a conversion script and related compatibility args to port Huggingface T5v1_1 weights to NeMo-Megatron. **Collection**: NLP # Changelog - Adds a state...
# What does this PR do ? Adds missing configs to GPT pre-training. **Collection**: NLP # Changelog - Upates the GPT pretraining yaml config and propages the args through to...
Signed-off-by: MaximumEntropy
I've added a theano cuDNN RNN implementation.
# What does this PR do ? Makes the changes necessary to do prompt learning of T5v1.1-converted checkpoints from HF. **Collection**: NLP # Changelog - Add specific line by line...