Trained with the dlm-cot pipeline.

Each step-* revision is a git branch containing the model weights (config.json, generation_config.json, model.safetensors, tokenizer.json, tokenizer_config.json) at that training step.