Lightning-AI/litgptGitHub
PyTorch/Lightning LLM pretraining/finetuning stack; pick if your stack is PyTorch, not PaddlePaddle.
20+ high-performance LLMs with recipes to pretrain, finetune and deploy at scale.
- 13.7k
- Python
- Apache-2.0
active29d ago
PaddlePaddle/PaddleFormers Alternatives
LLM/VLM training and model zoo on PaddlePaddle
PaddleFormers is a PaddlePaddle-based Transformers library that provides pretrained LLM/VLM model interfaces and supports large-scale distributed training with tensor, pipeline, and expert parallelism.
Same job, different approach
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
NVIDIA/Megatron-LMGitHub
PyTorch research framework for distributed transformer training; choose it for Megatron parallelism on Nvidia GPUs.
Ongoing research training transformer models at scale
- 17.7k
- Python
- NOASSERTION
active1mo ago
modelscope/ms-swiftGitHub
PyTorch/ModelScope suite for SFT/DPO/GRPO across many LLMs; choose for non-Paddle training workflows.
Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, Phi4, ...) (AAAI 2025).
- 15.7k
- Python
- Apache-2.0
active22d ago
NVIDIA-NeMo/AutomodelGitHub
PyTorch-native distributed LLM/VLM training with Hugging Face-friendly interface; choose for PyTorch basis.
🚀 Pytorch Distributed native training library for LLMs/VLMs with OOTB Hugging Face support
- 963
- Python
- Apache-2.0
active22d ago
AI-Hypercomputer/maxtextGitHub
JAX-based scalable LLM training library; choose for JAX/TPU stacks rather than Paddle.
A simple, performant, and scalable Jax LLM!
- 2.4k
- Python
- Apache-2.0
active22d ago
hiyouga/LlamaFactoryGitHub
PyTorch/HF fine-tuning framework for 100+ LLMs; choose for PEFT/fine-tuning workflows outside Paddle.
Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)
- 74.9k
- Python
- Apache-2.0
active29d ago
Older generation, or narrower
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
jingyaogong/minimindGitHub
Educational from-scratch training for a single tiny 64M model; valid only for small/learning scenarios.
🧠 Train a 64M-parameter LLM from scratch in just 2h!
- 61.7k
- Python
- Apache-2.0
active25d ago
DevJadhav/deepseek-from-scratchGitHub
From-scratch implementation of a single DeepSeek family; choose only for studying that specific model.
production ready deepseek R1 and v3.2 implementation in rust and python programming from scratch
- 202
- Python
- Apache-2.0
slowing9mo ago
DLLXW/baby-llama2-chineseGitHub
Small Chinese LLaMA2 pretrain/SFT recipe for a single 24G GPU; educational narrow scope.
用于从头预训练+SFT一个小参数量的中文LLaMa2的仓库；24G单卡即可运行得到一个具备简单中文问答能力的chat-llama2.
- 2.9k
- Python
- MIT
active2mo ago
Listed as an alternative to
These projects were analysed and named PaddleFormers among their alternatives. The relationship is not symmetric — how PaddleFormers rates them is a separate judgement, made when PaddleFormers is analysed in its own right.
NVIDIA/Megatron-LMGitHub
calls PaddleFormers “PaddlePaddle-based transformer training library for users committed to that framework.”
Ongoing research training transformer models at scale
17.7kPythonactive