---
license: apache-2.0
base_model: Qwen/Qwen3-8B
pipeline_tag: text-generation
---

# Qwen3-8B-A2D-untrained-dllm-convert

This repository contains the untrained initialization of Qwen3-8B converted to the A2D architecture (bidirectional attention), as introduced in the paper [Data-Efficient Autoregressive-to-Diffusion Language Models via On-Policy Distillation](https://huggingface.co/papers/2606.06712).

- **Project Page:** [https://opdlm.vercel.app/](https://opdlm.vercel.app/)
- **Code:** [https://github.com/divelab/OPDLM](https://github.com/divelab/OPDLM)
- **Base model:** [Qwen/Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B)

## Model Details

Qwen3-8B converted to A2D architecture (bidirectional attention) using [dllm convert pipeline](https://github.com/ZHZisZZ/dllm/blob/b8d76ff74b2053d359cd88fedfbc6362db17e3d7/examples/a2d/README.md?plain=1#L49-L53).

- **Architecture**: A2D-Qwen3 (non-causal attention, same weights as original)
- **Parameters**: 8.19B
- **Vocab size**: 151936
- **Model type**: `a2d-qwen3`

This model has the original Qwen3-8B weights with bidirectional (non-causal) attention. No diffusion pretraining or SFT has been applied.

**Mask token registration**: The mask token `<|MASK|>` (ID 151669) is registered in the tokenizer for use with diffusion-based language modeling. The original Qwen3 tokenizer includes `<|MASK|>` in `special_tokens_map.json` but does not register it in `tokenizer_config.json`, so `tokenizer.mask_token_id` returns `None`. We fixed this by adding `<|MASK|>` to the `added_tokens_decoder` section and the `mask_token` field in `tokenizer_config.json`, and adding the full `mask_token` entry in `special_tokens_map.json`. After this fix, `tokenizer.mask_token_id` correctly returns `151669`.