NeMo Framework Megatron Backend

NVIDIA

Container

NVIDIA

NeMo Framework Megatron Backend

NVIDIA NeMo™ framework Megatron backend supports pre-training, post-training, and reinforcement learning of LLMs and multi-modal generative AI models with state-of-the-art data processing, model training techniques, and flexible deployment options.

NVIDIA AI Enterprise Supported

Layer	Label		Created
38e2e19de1d9783d052cc23be0a22e8fccc9f6bed29269799ede76d91f8c1d0c	CONFIG	`Entrypoint /opt/nvidia/nvidia_entrypoint.sh; WorkingDir /workspace; ExposedPorts 6006/tcp, 8888/tcp`	04/07/2025 8:27 AM UTC
1f0f1ad53953ad5b9c79243b300debb517f43b73da772bf2a14a5840db8cedb2	ARG	`NVIDIA_BUILD_REF`	04/07/2025 8:27 AM UTC
6b3a57b8399c12cabee59faf2c76117bfad967c1d6bb69ef5ada80a6aebf0a5f	LABEL	`com.nvidia.build.id=148941828`	04/07/2025 8:27 AM UTC
983b11c576de36f4fb2a072fbd6ab2ec2a242e81d35ddb281cec4f03eb5cc5e7	ENV	`NVIDIA_BUILD_ID=148941828`	04/07/2025 8:27 AM UTC
c96ec1ec00adb206e4eff00a1d398c72490a273b710b95c29743f441d0e63020	ARG	`NVIDIA_BUILD_ID`	04/07/2025 8:27 AM UTC
b708a5703ea002825755d3764283629a9e255611c36c406048d049c157c66955	ENV	`NEMO_FW_PIPELINE_ID=26557816`	04/07/2025 8:27 AM UTC
007055a3957d0ee98ea26797291bee0fa8bf700c5717d66af40bbd15c59789e1	ARG	`NEMO_FW_PIPELINE_ID=26557816`	04/07/2025 8:27 AM UTC
659a2fbbfc8d47dd58f264588e5e15b1676186a1a8f301f2b468d6e30fbf181d	RUN	NVIDIA_BIGNLP_VERSION= TRT_VER=10.9.0.34 CUDA_VER=12.8 CUDNN_VER=9.8.0.87-1 NCCL_VER=2.25.1-1+cuda12.8 CUBLAS_VER=12.8.4.1-1 NVRTC_VER=12.8.93-1 TRT_LLM_COMMIT=fe85b725df3e769b87a8fdb63154276f87032080 TRT_LLM_VERSION=0.17.1 REINSTALL_NSYS=True NSYS_VERSION=2025.1.1.65 REINSTALL_APEX=False APEX_COMMIT=379937e1719b8190f38f3e666e2a8d32c97e221f MCORE_COMMIT=0d5bee6b99c3191901e7f733a85ca8e4bd158994 MCORE_VERSION=0.12.0rc0 SRC_DIR=/src/tensorrt_llm CPP_BUILD_DIR=/src/tensorrt_llm/cpp/build TARGET_ARCH=x86 INSTALL_VLLM=True ALIGNER_COMMIT=dev ALIGNER_VERSION=0.6.0rc0.dev0 LAUNCHER_COMMIT=9b715db12f8c78fa33cfd2414e12d1dabfa8e8ad CURATOR_COMMIT=3169a24dcf75237c2c9eddf263c0ef0d62ac6138 INSTALL_CURATOR=True CURATOR_VERSION=0.8.0rc0.dev0 NEMO_COMMIT=95c9edead4c70a5e104d7bb046c5fb611f60c263 NEMO_VERSION=2.3.0rc0 INSTALL_RESIL=False RESIL_COMMIT=not_installed RESIL_VERSION=not_installed INSTALL_MODELOPT=False MODELOPT_VERSION= NEMO_RUN_COMMIT=93afc7a5a6297ad978730697da1f40c54fcafb5d PRE_COMPILE_QUANT_KERNELS=True chmod -R a+w /workspace	04/07/2025 8:27 AM UTC
0234567211f5f5bf83b08231005a7e30ba44cdbf917853b143754682ec04ec3d	WORKDIR	`/workspace`	04/07/2025 8:27 AM UTC
3c3d599ca5b605e4ecdc1a025722c08b51419fdc95e2793a70f72e5622658b68	RUN	NVIDIA_BIGNLP_VERSION= TRT_VER=10.9.0.34 CUDA_VER=12.8 CUDNN_VER=9.8.0.87-1 NCCL_VER=2.25.1-1+cuda12.8 CUBLAS_VER=12.8.4.1-1 NVRTC_VER=12.8.93-1 TRT_LLM_COMMIT=fe85b725df3e769b87a8fdb63154276f87032080 TRT_LLM_VERSION=0.17.1 REINSTALL_NSYS=True NSYS_VERSION=2025.1.1.65 REINSTALL_APEX=False APEX_COMMIT=379937e1719b8190f38f3e666e2a8d32c97e221f MCORE_COMMIT=0d5bee6b99c3191901e7f733a85ca8e4bd158994 MCORE_VERSION=0.12.0rc0 SRC_DIR=/src/tensorrt_llm CPP_BUILD_DIR=/src/tensorrt_llm/cpp/build TARGET_ARCH=x86 INSTALL_VLLM=True ALIGNER_COMMIT=dev ALIGNER_VERSION=0.6.0rc0.dev0 LAUNCHER_COMMIT=9b715db12f8c78fa33cfd2414e12d1dabfa8e8ad CURATOR_COMMIT=3169a24dcf75237c2c9eddf263c0ef0d62ac6138 INSTALL_CURATOR=True CURATOR_VERSION=0.8.0rc0.dev0 NEMO_COMMIT=95c9edead4c70a5e104d7bb046c5fb611f60c263 NEMO_VERSION=2.3.0rc0 INSTALL_RESIL=False RESIL_COMMIT=not_installed RESIL_VERSION=not_installed INSTALL_MODELOPT=False MODELOPT_VERSION= NEMO_RUN_COMMIT=93afc7a5a6297ad978730697da1f40c54fcafb5d PRE_COMPILE_QUANT_KERNELS=True if [ $PRE_COMPILE_QUANT_KERNELS = "True" ]; then python -c "import modelopt.torch.quantization.extensions as ext; ext.precompile()"; fi	04/07/2025 8:27 AM UTC