Mask BEFORE exp to avoid inf*0=NaN in bf16

Memory-efficient SSD scan: einsum contractions + bf16

Update README.md

Upload openNemo.jpg

Delete safety.md

Delete privacy.md

Delete bias.md

Delete explainability.md

Rename modeling_nemotron_h2.py to modeling_nemotron_h.py

Rename modeling_nemotron_h.py to modeling_nemotron_h2.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

Changed aliases into propertys to prevent crashes on PEFT

Fixed _update_model_kwargs_for_generation

Fixed attention_mask during gen

Update modeling_nemotron_h.py

Update modeling_nemotron_h.py

added Self Attn and MLP

Update configuration_nemotron_h.py

Update configuration_nemotron_h.py

Delete acc-vs-budget.png

Delete accuracy_chart.png

Update modeling_nemotron_h.py

Update configuration_nemotron_h.py

Update README.md

Update README.md

Update README.md

Duplicate from nvidia/NVIDIA-Nemotron-Nano-9B-v2