pytorch/torchtitan
A PyTorch native platform for training generative AI models
PyTorch-native stack that spans pretraining through RL — torchtitan is a clean-room, PyTorch-native codebase that takes a model from pretraining through SFT to reinforcement learning, composing FSDP2, tensor, pipeline, context, and expert parallelism without rewriting model code. TitanRL now shares the same model definitions and kernels between the trainer and vLLM generation, with a bitwise-reproducible mode, and it already supports DeepSeek, Qwen, Kimi, gpt-oss, Llama, and FLUX.
Metrics
Retention can be computed 30 days after a repository first trends (it isn't shown until then). See About for how each metric is defined.
Star growth and Trending appearances (last 30 days)
Bands mark days on the trending list (darker = higher rank)
Daily trending ranks as a table (1 days)
| Date | Daily |
|---|---|
| Nov 20, 2026 | python #13 0 |
Created Dec 13, 2023 · First on trending Nov 20, 2026