Qwen3-0.6B-diffusion-mdlm-v0.1

Qwen3-0.6B-diffusion-mdlm-v0.1 is a diffusion-based language model adapted from Qwen3-0.6B using MDLM (masked diffusion), trained with the dLLM framework.

Model Overview

Qwen3-0.6B-diffusion-mdlm-v0.1 has the following features:

• Method: Masked Diffusion Language Modeling (MDLM)
• Framework: dLLM
• Base Model: Qwen3-0.6B
• Datasets: tulu-3-sft-mixture, smoltalk, opc-sft-stage1 and opc-sft-stage2

For training details, see the W&B report.

Installation

pip install torch transformers accelerate

Quick Start

import torch
import numpy as np
import torch.nn.functional as F

from transformers import AutoTokenizer, AutoModelForMaskedLM


def add_gumbel_noise(logits, temperature):
    if temperature == 0:
        return logits
    logits = logits.to(torch.float64)
    noise = torch.rand_like(logits, dtype=torch.float64)
    gumbel_noise = (- torch.log(noise)) ** temperature
    return logits.exp() / gumbel_noise


def get_num_transfer_tokens(mask_index, steps):
    mask_num = mask_index.sum(dim=1, keepdim=True)
    base = mask_num // steps
    remainder = mask_num % steps
    num_transfer_tokens = torch.zeros(mask_num.size(0), steps, device=mask_index.device, dtype=torch.int64) + base
    for i in range(mask_num.size(0)):
        num_transfer_tokens[i, :remainder[i]] += 1
    return num_transfer_tokens


@torch.no_grad()
def generate(model, prompt, prompt_lens, pad_id, steps=128, max_new_tokens=128, block_size=64, temperature=0.0, cfg_scale=0.0, remasking="random"):
    mask_id = tokenizer.mask_token_id
    batch_size = prompt.size(0)
    total_length = int(prompt_lens.max().item() + max_new_tokens)
    x = torch.full((batch_size, total_length), pad_id, dtype=torch.long, device=model.device)
    for i, length in enumerate(prompt_lens.tolist()):
        x[i, :length] = prompt[i, :length]
        x[i, length : length + max_new_tokens] = mask_id

    prompt_index = torch.arange(total_length, device=x.device).unsqueeze(0) < prompt_lens.unsqueeze(1)
    positions = torch.arange(total_length, device=x.device)

    assert max_new_tokens % block_size == 0
    num_blocks = max_new_tokens // block_size
    assert steps % num_blocks == 0
    steps_per_block = steps // num_blocks

    for num_block in range(num_blocks):
        block_start = prompt_lens + num_block * block_size
        block_end = block_start + block_size
        init_block_mask = (
            (positions.unsqueeze(0) >= block_start.unsqueeze(1))
            & (positions.unsqueeze(0) < block_end.unsqueeze(1))
            & (x == mask_id)
        )
        num_transfer_tokens = get_num_transfer_tokens(init_block_mask, steps_per_block)

        for i in range(steps_per_block):
            block_mask = (
                (positions.unsqueeze(0) >= block_start.unsqueeze(1))
                & (positions.unsqueeze(0) < block_end.unsqueeze(1))
                & (x == mask_id)
            )

            if cfg_scale > 0.0:
                un_x = x.clone()
                un_x[prompt_index] = mask_id
                x_ = torch.cat([x, un_x], dim=0)
                logits = model(x_).logits
                logits, un_logits = torch.chunk(logits, 2, dim=0)
                logits = un_logits + (cfg_scale + 1.0) * (logits - un_logits)
            else:
                logits = model(x).logits

            logits_with_noise = add_gumbel_noise(logits, temperature=temperature)
            x0 = torch.argmax(logits_with_noise, dim=-1)

            if remasking == "low_confidence":
                p = F.softmax(logits, dim=-1)
                x0_p = torch.gather(p, dim=-1, index=x0.unsqueeze(-1)).squeeze(-1)
            elif remasking == "random":
                x0_p = torch.rand_like(x0, dtype=torch.float)
            else:
                raise NotImplementedError(remasking)

            confidence = torch.full_like(x0_p, -np.inf)
            confidence = torch.where(block_mask, x0_p, confidence)

            x0 = torch.where(block_mask, x0, x)

            transfer_index = torch.zeros_like(x0, dtype=torch.bool, device=x0.device)
            for j in range(confidence.shape[0]):
                k = int(num_transfer_tokens[j, i].item())
                if k == 0:
                    continue
                _, select_index = torch.topk(confidence[j], k=k)
                transfer_index[j, select_index] = True
            x[transfer_index] = x0[transfer_index]

    return x

device = "cuda" if torch.cuda.is_available() else "cpu"
model = AutoModelForMaskedLM.from_pretrained("dllm-collection/Qwen3-0.6B-diffusion-mdlm-v0.1", dtype=torch.bfloat16, trust_remote_code=True).to(device).eval()
tokenizer = AutoTokenizer.from_pretrained("dllm-collection/Qwen3-0.6B-diffusion-mdlm-v0.1")
if tokenizer.pad_token_id is None and tokenizer.eos_token is not None:
    tokenizer.pad_token = tokenizer.eos_token
pad_id = tokenizer.pad_token_id or tokenizer.eos_token_id or tokenizer.mask_token_id

messages = [
    [
        {"role": "system", "content": "You are a helpful AI assistant."},
        {"role": "user", "content": "Implement a DFS traversal in Python with clear inline comments."},
    ],
    [
        {"role": "system", "content": "You are a helpful AI assistant."},
        {"role": "user", "content": "Lily can run 12 kilometers per hour for 4 hours. After that, she runs 10 kilometers per hour. How many kilometers can she run in 10 hours?"},
    ],
]

encoded = [tokenizer.apply_chat_template(m, add_generation_prompt=True, tokenize=True) for m in messages]
prompt_lens = torch.tensor([len(e) for e in encoded], dtype=torch.long)
max_prompt_len = max(prompt_lens).item()
prompt_tensor = torch.full((len(encoded), max_prompt_len), pad_id, dtype=torch.long)
for i, ids in enumerate(encoded):
    prompt_tensor[i, : len(ids)] = torch.tensor(ids, dtype=torch.long)

prompt_tensor = prompt_tensor.to(device)
prompt_lens = prompt_lens.to(device)
max_new_tokens = 256

text = generate(
    model, prompt_tensor, prompt_lens, pad_id=pad_id, steps=256, max_new_tokens=max_new_tokens, block_size=64, temperature=0.0, cfg_scale=0.0, remasking="low_confidence"
)

new_tokens = [
    text[i, prompt_lens[i] : prompt_lens[i] + max_new_tokens].tolist() for i in range(text.size(0))
]
for idx, decoded in enumerate(tokenizer.batch_decode(new_tokens, skip_special_tokens=False)):
    print(f"\n[Sample {idx}]")
    print(decoded)

Generation Parameters

Parameter	Description	Default
max_new_tokens	Number of tokens to generate	256
steps	Number of diffusion denoising iterations	256
temperature	Sampling temperature; set to 0.0 for deterministic generation	0.0
block_size	Token block size used during iterative denoising	64
cfg_scale	Classifier-free guidance scale controlling instruction adherence (higher = more deterministic)	0.0
remasking	Strategy for re-masking during each denoising step (random or low_confidence)	low_confidence

Command-Line Interface

Follow the Github repo's demo script examples/a2d/mdlm/chat.py for visualized generation:

python -u examples/a2d/bd3lm/chat.py \
    --model_name_or_path dllm-collection/Qwen3-0.6B-diffusion-bd3lm-v0.1 \
    --chat_template True --block_size 64 --remasking low_confidence --steps 256 --max_new_tokens 256

Evaluation

Model	GSM8K	MATH	BBH	MMLU‑Pro	Hellaswag	MMLU	HumanEval	MBPP
Qwen3-0.6B-diffusion-bd3lm-v0.1 (evaluated)	46.6	13.9	27.0	14.1	40.0	38.8	47.6	32.0
Qwen3-0.6B-diffusion-mdlm-v0.1 (evaluated)	29.8	8.8	27.0	17.6	42.1	40.0	30.5	29.2
Qwen3-0.6B (reported)	59.6	32.4	41.5	24.7	47.4	52.8	32.3	36.6
Qwen2.5-0.5B (reported)	41.6	19.5	20.3	15.7	52.1	47.5	30.5	39.3

To automatically evaluate Qwen3-0.6B-diffusion-mdlm-v0.1 on all benchmarks, run:

bash examples/a2d/mdlm/eval.sh \
  --model_name_or_path dllm-collection/Qwen3-0.6B-diffusion-mdlm-v0.1

Citation

If you use Qwen3-0.6B-diffusion-mdlm-v0.1 or dLLM, please cite:

@misc{dllm,
  author = {Zhanhui Zhou and Lingjie Chen and Hanghang Tong and Dawn Song},
  title = {dLLM: Simple Diffusion Language Modeling},
  year = {2025},
  howpublished = {\url{https://github.com/ZHZisZZ/dllm}},
}