Qwen3-4B-DFlash-b16

Qwen3-4B-DFlash-b16 is a diffusion draft model trained against a frozen Qwen3 target for speculative decoding.

Train Qwen3-4B-DFlash-b16

From the repository root, run:

uv run torchrun --nproc-per-node=8 nemo_automodel/recipes/dllm/train_ft.py \
  -c examples/dllm_sft/dflash_sft.yaml

Model Reference

Model Architecture

Example Model and Recipe

See the dLLM Fine-Tuning Guide.