home/categories/framework-internals/orchestra-research-ai-research-skills-01-model-architecture-torchtitan-skill-md
framework-internalsdevelopment
distributed-llm-pretraining-torchtitan
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
maintainer
Orchestra-Research
Actualizado 1/29/2026
Estrellas
6563
Forks
515
quick start
Installation and usage
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
Instalación
$ install --globalskills.sh
Uso
Después de instalarlo, puedes usar este skill ejecutando el siguiente comando en tu terminal:
skills use distributed-llm-pretraining-torchtitan