Entrypoint /opt/nvidia/nvidia_entrypoint.sh; WorkingDir /opt/NeMo-Curator; ExposedPorts 6006/tcp, 8888/tcp

python -c "from transformers import AutoTokenizer; tok_llama3_8b=AutoTokenizer.from_pretrained('meta-llama/Meta-Llama-3-8B');  tok_llama3_70b=AutoTokenizer.from_pretrained('meta-llama/Meta-Llama-3-70B');  tok_llama3_8b_inst=AutoTokenizer.from_pretrained('meta-llama/Meta-Llama-3-8B-instruct');  tok_llama3_70b_inst=AutoTokenizer.from_pretrained('meta-llama/Meta-Llama-3-70B-instruct');"

RUN /bin/sh -c <<EOFINNER bash -eux rsync -ahP /mnt/tokenizer-dir/ ~/.cache/huggingface/hub/ EOFINNER # buildkit

python -c "from transformers import AutoTokenizer; tok_llama3_8b=AutoTokenizer.from_pretrained('meta-llama/Meta-Llama-3-8B');  tok_llama3_70b=AutoTokenizer.from_pretrained('meta-llama/Meta-Llama-3-70B');  tok_llama3_8b_inst=AutoTokenizer.from_pretrained('meta-llama/Meta-Llama-3-8B-instruct');  tok_llama3_70b_inst=AutoTokenizer.from_pretrained('meta-llama/Meta-Llama-3-70B-instruct');"

RUN /bin/sh -c <<EOFINNER bash -eux cd /mnt/tokenizer-dir mkdir -p /tmp_assets tar -xzvf llama3_8b_tokenizer.tar.gz -C /tmp_assets tar -xzvf llama3_70b_tokenizer.tar.gz -C /tmp_assets cp -rf /tmp_assets/hub ~/.cache/huggingface rm -rf /tmp_assets EOFINNER # buildkit

RUN /bin/sh -c <<EOFINNER bash -eux git config --global user.email "nemo@nvidia.com" git config --global user.name "NeMo Framework" pushd /opt/NeMo-Aligner git am --3way <./aligner-llama3.1.patch git am --3way <./aligner-calculator.patch rm -f ./aligner-llama3.1.patch ./aligner-calculator.patch popd pushd /opt/megatron-lm git am --3way <./mcore-calculator.patch rm -f ./mcore-calculator.patch popd git config --global --unset user.email git config --global --unset user.name EOFINNER # buildkit

././mcore-calculator.patch /opt/megatron-lm

././aligner-calculator.patch /opt/NeMo-Aligner

././aligner-llama3.1.patch /opt/NeMo-Aligner

org.opencontainers.image.title=nvcr.io/nvidian/nemo:24.05.llama3.1_nc