All Tools
E
Fine-tuningFreeOpen Source
EASYLM
JAX/Flax pretrain, finetune, eval, and serve for LLMs
Apache-2.0
ABOUT
JAX LLM work usually means gluing training, checkpoint conversion, eval, and serving yourself. EasyLM packages those steps for LLaMA-class models so a small team can pretrain or finetune in Flax and serve the result without a Megatron or Hugging Face Trainer rewrite.
INTEGRATION GUIDE
1. Finetune a LLaMA-class model in JAX/Flax with published scripts
2. Convert and serve a JAX checkpoint after pretraining
3. Evaluate generated text without leaving the EasyLM workflow
4. Prototype JAX training on TPU or GPU before moving to MaxText
TAGS
pythonjaxflaxllamatrainingservingopen-source