vllm-omni/docs/user_guide/diffusion/parallelism_acceleration.md

# Parallelism Acceleration Guide

This guide includes how to use parallelism methods in vLLM-Omni to speed up diffusion model inference as well as reduce the memory requirement on each device.

## Overview

The following parallelism methods are currently supported in vLLM-Omni:

1. DeepSpeed Ulysses Sequence Parallel (DeepSpeed Ulysses-SP) ([arxiv paper](https://arxiv.org/pdf/2309.14509)): Ulysses-SP splits the input along the sequence dimension and uses all-to-all communication to allow each device to compute only a subset of attention heads.

2. [Ring-Attention](#ring-attention) - splits the input along the sequence dimension and uses ring-based P2P communication to accumulate attention results, keeping the sequence dimension sharded

3. Classifier-Free-Guidance Parallel (CFG-Parallel): CFG-Parallel runs the positive/negative prompts of classifier-free guidance (CFG) on different devices, then merges on a single device to perform the scheduler step.

4. [Tensor Parallelism](#tensor-parallelism): Tensor parallelism shards model weights across devices. This can reduce per-GPU memory usage. Note that for diffusion models we currently shard the majority of layers within the DiT.

The following table shows which models are currently supported by parallelism method:

### ImageGen

| Model                    | Model Identifier                     | Ulysses-SP | Ring-SP | CFG-Parallel | Tensor-Parallel |
|--------------------------|--------------------------------------|:----------:|:-------:|:------------:|:---------------:|
| **LongCat-Image**        | `meituan-longcat/LongCat-Image`      |     ✅      |    ✅    |      ❌       |        ✅        |
| **LongCat-Image-Edit**   | `meituan-longcat/LongCat-Image-Edit` |     ✅      |    ✅    |      ❌       |        ✅        |
| **Ovis-Image**           | `OvisAI/Ovis-Image`                  |     ❌      |    ❌    |      ❌       |        ❌        |
| **Qwen-Image**           | `Qwen/Qwen-Image`                    |     ✅      |    ✅    |      ✅       |        ✅        |
| **Qwen-Image-Edit**      | `Qwen/Qwen-Image-Edit`               |     ✅      |    ✅    |      ✅       |        ✅        |
| **Qwen-Image-Edit-2509** | `Qwen/Qwen-Image-Edit-2509`          |     ✅      |    ✅    |      ✅       |        ✅        |
| **Qwen-Image-Layered**   | `Qwen/Qwen-Image-Layered`            |     ✅      |    ✅    |      ✅       |        ✅        |
| **Z-Image**              | `Tongyi-MAI/Z-Image-Turbo`           |     ✅      |    ✅    |      ❌       |  ✅ (TP=2 only)  |
| **Stable-Diffusion3.5**  | `stabilityai/stable-diffusion-3.5`   |     ❌      |    ❌    |      ❌       |        ❌        |
| **FLUX.2-klein**         | `black-forest-labs/FLUX.2-klein-4B`  |     ❌      |    ❌    |      ❌       |        ✅        |
| **FLUX.1-dev**           | `black-forest-labs/FLUX.1-dev`       |     ❌      |    ❌    |      ❌       |        ✅        |

!!! note "TP Limitations for Diffusion Models"
    We currently implement Tensor Parallelism (TP) only for the DiT (Diffusion Transformer) blocks. This is because the `text_encoder` component in vLLM-Omni uses the original Transformers implementation, which does not yet support TP.

    - Good news: The text_encoder typically has minimal impact on overall inference performance.
    - Bad news: When TP is enabled, every TP process retains a full copy of the text_encoder weights, leading to significant GPU memory waste.

    We are actively refactoring this design to address this. For details and progress, please refer to [Issue #771](https://github.com/vllm-project/vllm-omni/issues/771).


!!! note "Why Z-Image is TP=2 only"
    Z-Image Turbo is currently limited to `tensor_parallel_size` of **1 or 2** due to model shape divisibility constraints.
    For example, the model has `n_heads=30` and a final projection out dimension of `64`, so valid TP sizes must divide both 30 and 64; the only common divisors are **1 and 2**.

### VideoGen

| Model | Model Identifier | Ulysses-SP | Ring-SP | Tensor-Parallel |
|-------|------------------|------------|---------|--------------------------|
| **Wan2.2** | `Wan-AI/Wan2.2-T2V-A14B-Diffusers` | ✅ | ✅ | ✅ |

### Tensor Parallelism

Tensor parallelism splits model parameters across GPUs. In vLLM-Omni, tensor parallelism is configured via `DiffusionParallelConfig.tensor_parallel_size`.

#### Offline Inference

```python
from vllm_omni import Omni
from vllm_omni.diffusion.data import DiffusionParallelConfig

omni = Omni(
    model="Tongyi-MAI/Z-Image-Turbo",
    parallel_config=DiffusionParallelConfig(tensor_parallel_size=2),
)

outputs = omni.generate(
    "a cat reading a book",
    OmniDiffusionSamplingParams(
        num_inference_steps=9,
        width=512,
        height=512,
    ),
)
```

### Sequence Parallelism

#### Ulysses-SP

##### Offline Inference

An example of offline inference script using [Ulysses-SP](https://arxiv.org/pdf/2309.14509) is shown below:
```python
from vllm_omni import Omni
from vllm_omni.inputs.data import OmniDiffusionSamplingParams
from vllm_omni.diffusion.data import DiffusionParallelConfig
ulysses_degree = 2

omni = Omni(
    model="Qwen/Qwen-Image",
    parallel_config=DiffusionParallelConfig(ulysses_degree=2)
)

outputs = omni.generate(
    "A cat sitting on a windowsill",
    OmniDiffusionSamplingParams(num_inference_steps=50, width=2048, height=2048),
)
```

See `examples/offline_inference/text_to_image/text_to_image.py` for a complete working example.

##### Online Serving

You can enable Ulysses-SP in online serving for diffusion models via `--usp`:

```bash
# Text-to-image (requires >= 2 GPUs)
vllm serve Qwen/Qwen-Image --omni --port 8091 --usp 2
```

##### Benchmarks
!!! note "Benchmark Disclaimer"
    These benchmarks are provided for **general reference only**. The configurations shown use default or common parameter settings and have not been exhaustively optimized for maximum performance. Actual performance may vary based on:

    - Specific model and use case
    - Hardware configuration
    - Careful parameter tuning
    - Different inference settings (e.g., number of steps, image resolution)


To measure the parallelism methods, we run benchmarks with **Qwen/Qwen-Image** model generating images (**2048x2048** as long sequence input) with 50 inference steps. The hardware devices are NVIDIA H800 GPUs. `sdpa` is the attention backends.

| Configuration | Ulysses degree |Generation Time | Speedup |
|---------------|----------------|---------|---------|
| **Baseline (diffusers)** | - | 112.5s | 1.0x |
| Ulysses-SP  |  2  |  65.2s | 1.73x |
| Ulysses-SP  |  4  | 39.6s | 2.84x |
| Ulysses-SP  |  8  | 30.8s | 3.65x |

#### Ring-Attention

Ring-Attention ([arxiv paper](https://arxiv.org/abs/2310.01889)) splits the input along the sequence dimension and uses ring-based P2P communication to accumulate attention results. Unlike Ulysses-SP which uses all-to-all communication, Ring-Attention keeps the sequence dimension sharded throughout the computation and circulates Key/Value blocks through a ring topology.

##### Offline Inference

An example of offline inference script using Ring-Attention is shown below:
```python
from vllm_omni import Omni
from vllm_omni.inputs.data import OmniDiffusionSamplingParams
from vllm_omni.diffusion.data import DiffusionParallelConfig
ring_degree = 2

omni = Omni(
    model="Qwen/Qwen-Image",
    parallel_config=DiffusionParallelConfig(ring_degree=2)
)

outputs = omni.generate(
    "A cat sitting on a windowsill",
    OmniDiffusionSamplingParams(num_inference_steps=50, width=2048, height=2048),
)
```

See `examples/offline_inference/text_to_image/text_to_image.py` for a complete working example.


##### Online Serving

You can enable Ring-Attention in online serving for diffusion models via `--ring`:

```bash
# Text-to-image (requires >= 2 GPUs)
vllm serve Qwen/Qwen-Image --omni --port 8091 --ring 2
```

##### Benchmarks
!!! note "Benchmark Disclaimer"
    These benchmarks are provided for **general reference only**. The configurations shown use default or common parameter settings and have not been exhaustively optimized for maximum performance. Actual performance may vary based on:

    - Specific model and use case
    - Hardware configuration
    - Careful parameter tuning
    - Different inference settings (e.g., number of steps, image resolution)


To measure the parallelism methods, we run benchmarks with **Qwen/Qwen-Image** model generating images (**1024x1024** as long sequence input) with 50 inference steps. The hardware devices are NVIDIA A100 GPUs. `flash_attn` is the attention backends.

| Configuration | Ring degree |Generation Time | Speedup |
|---------------|----------------|---------|---------|
| **Baseline (diffusers)** | - | 45.2s | 1.0x |
| Ring-Attention  |  2  |  29.9s | 1.51x |
| Ring-Attention  |  4  | 23.3s | 1.94x |


#### Hybrid Ulysses + Ring

You can combine both Ulysses-SP and Ring-Attention for larger scale parallelism. The total sequence parallel size equals `ulysses_degree × ring_degree`.

##### Offline Inference

```python
from vllm_omni import Omni
from vllm_omni.inputs.data import OmniDiffusionSamplingParams
from vllm_omni.diffusion.data import DiffusionParallelConfig

# Hybrid: 2 Ulysses × 2 Ring = 4 GPUs total
omni = Omni(
    model="Qwen/Qwen-Image",
    parallel_config=DiffusionParallelConfig(ulysses_degree=2, ring_degree=2)
)

outputs = omni.generate(
    "A cat sitting on a windowsill",
    OmniDiffusionSamplingParams(num_inference_steps=50, width=2048, height=2048),
)
```

##### Online Serving

```bash
# Text-to-image (requires >= 4 GPUs)
vllm serve Qwen/Qwen-Image --omni --port 8091 --usp 2 --ring 2
```

##### Benchmarks
!!! note "Benchmark Disclaimer"
    These benchmarks are provided for **general reference only**. The configurations shown use default or common parameter settings and have not been exhaustively optimized for maximum performance. Actual performance may vary based on:

    - Specific model and use case
    - Hardware configuration
    - Careful parameter tuning
    - Different inference settings (e.g., number of steps, image resolution)


To measure the parallelism methods, we run benchmarks with **Qwen/Qwen-Image** model generating images (**1024x1024** as long sequence input) with 50 inference steps. The hardware devices are NVIDIA A100 GPUs. `flash_attn` is the attention backends.

| Configuration | Ulysses degree | Ring degree | Generation Time | Speedup |
|---------------|----------------|-------------|-----------------|---------|
| **Baseline (diffusers)** | - | - | 45.2s | 1.0x |
| Hybrid Ulysses + Ring  |  2  |  2  |  24.3s | 1.87x |


##### How to parallelize a new model

NOTE: "Terminology: SP vs CP"
    Our "Sequence Parallelism" (SP) corresponds to "Context Parallelism" (CP) in the [diffusers library](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/_modeling_parallel.py).
    We use "Sequence Parallelism" to align with vLLM-Omni's terminology.

---

###### Non-intrusive `_sp_plan` (Recommended)

The `_sp_plan` mechanism allows SP without modifying `forward()` logic. The framework automatically registers hooks to shard inputs and gather outputs at module boundaries.

**Requirements for `forward()` function:**

- Tensor operations that need sharding/gathering must happen at **`nn.Module` boundaries** (not inline Python operations)
- If your `forward()` contains inline tensor operations (e.g., `torch.cat`, `pad_sequence`) that need sharding, **extract them into a submodule**

**When to create a submodule:**

```python
# ❌ BAD: Inline operations - hooks cannot intercept
def forward(self, x, cap_feats):
    unified = torch.cat([x, cap_feats], dim=1)  # Cannot be sharded via _sp_plan
    ...

# ✅ GOOD: Extract into a submodule
class UnifiedPrepare(nn.Module):
    def forward(self, x, cap_feats):
        return torch.cat([x, cap_feats], dim=1)  # Now can be sharded via _sp_plan

class MyModel(nn.Module):
    def __init__(self):
        self.unified_prepare = UnifiedPrepare()  # Submodule

    def forward(self, x, cap_feats):
        unified = self.unified_prepare(x, cap_feats)  # Hook can intercept here
```

---

###### Defining `_sp_plan`

**Type definitions** (see [diffusers `_modeling_parallel.py`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/_modeling_parallel.py) for reference):

```python
from vllm_omni.diffusion.distributed.sp_plan import (
    SequenceParallelInput,   # Corresponds to diffusers' ContextParallelInput
    SequenceParallelOutput,  # Corresponds to diffusers' ContextParallelOutput
)
```

| Parameter | Description |
|-----------|-------------|
| `split_dim` | Dimension to split/gather (usually `1` for sequence) |
| `expected_dims` | Expected tensor rank for validation (optional) |
| `split_output` | `False`: shard **input** parameters; `True`: shard **output** tensors |
| `auto_pad` | Auto-pad if sequence not divisible by world_size (Ulysses only) |

**Key naming convention:**

| Key | Meaning | Python equivalent |
|-----|---------|-------------------|
| `""` | Root model | `model` |
| `"blocks.0"` | First element of ModuleList | `model.blocks[0]` |
| `"blocks.*"` | All elements of ModuleList | `for b in model.blocks` |
| `"outputs.main"` | ModuleDict entry | `model.outputs["main"]` |

**Dictionary key types:**

| Key type | `split_output` | Description |
|----------|----------------|-------------|
| `"param_name"` (str) | `False` | Shard **input parameter** by name |
| `0`, `1` (int) | `True` | Shard **output tuple** by index |

**Example** (similar to [diffusers `transformer_wan.py`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/transformers/transformer_wan.py)):

```python
class MyTransformer(nn.Module):
    _sp_plan = {
        # Shard rope module OUTPUTS (returns tuple)
        "rope": {
            0: SequenceParallelInput(split_dim=1, expected_dims=4, split_output=True),  # cos
            1: SequenceParallelInput(split_dim=1, expected_dims=4, split_output=True),  # sin
        },
        # Shard transformer block INPUT parameter
        "blocks.0": {
            "hidden_states": SequenceParallelInput(split_dim=1, expected_dims=3),
        },
        # Gather at final projection
        "proj_out": SequenceParallelOutput(gather_dim=1, expected_dims=3),
    }
```

---

###### Hook flow

```
Input → [SequenceParallelSplitHook: pre_forward] → Module.forward() → [post_forward] → ...
                                                                              ↓
... → [SequenceParallelGatherHook: post_forward] → Output
```

1. **SplitHook** shards tensors before/after the target module
2. **Attention layers** handle Ulysses/Ring communication internally
3. **GatherHook** collects sharded outputs

The framework automatically applies these hooks when `sequence_parallel_size > 1`.

---

###### Method 2: Intrusive modification (For complex cases)

For models with dynamic sharding logic that cannot be expressed via `_sp_plan`:

```python
from vllm_omni.diffusion.distributed.sp_sharding import sp_shard, sp_gather

def forward(self, hidden_states, ...):
    if self.parallel_config.sequence_parallel_size > 1:
        hidden_states = sp_shard(hidden_states, dim=1)
        # ... computation ...
        output = sp_gather(output, dim=1)
    return output
```

---

###### Choosing the right approach

| Scenario | Approach |
|----------|----------|
| Standard transformer | `_sp_plan` |
| Inline tensor ops need sharding | Extract to submodule + `_sp_plan` |
| Dynamic/conditional sharding | Intrusive modification |


### CFG-Parallel

#### Offline Inference

CFG-Parallel is enabled through `DiffusionParallelConfig(cfg_parallel_size=2)`, which runs one rank for the positive branch and one rank for the negative branch.

An example of offline inference using CFG-Parallel (image-to-image) is shown below:

```python
from vllm_omni import Omni
from vllm_omni.diffusion.data import DiffusionParallelConfig

image_path = "path_to_image.png"
omni = Omni(
    model="Qwen/Qwen-Image-Edit",
    parallel_config=DiffusionParallelConfig(cfg_parallel_size=2),
)
input_image = Image.open(image_path).convert("RGB")

outputs = omni.generate(
    {
        "prompt": "turn this cat to a dog",
        "negative_prompt": "low quality, blurry",
        "multi_modal_data": {"image": input_image},
    },
    OmniDiffusionSamplingParams(
        true_cfg_scale=4.0,
        num_inference_steps=50,
    ),
)
```

Notes:

- CFG-Parallel is only effective when a `negative_prompt` is provided AND a guidance scale (or `cfg_scale`) is greater than 1.

See `examples/offline_inference/image_to_image/image_edit.py` for a complete working example.
```bash
cd examples/offline_inference/image_to_image/
python image_edit.py \
  --model "Qwen/Qwen-Image-Edit" \
  --image "qwen_image_output.png" \
  --prompt "turn this cat to a dog" \
  --negative_prompt "low quality, blurry" \
  --cfg_scale 4.0 \
  --output "edited_image.png" \
  --cfg_parallel_size 2
```

#### Online Serving

You can enable CFG-Parallel in online serving for diffusion models via `--cfg-parallel-size`:

```bash
vllm serve Qwen/Qwen-Image-Edit --omni --port 8091 --cfg-parallel-size 2
```

#### How to parallelize a pipeline

This section describes how to add CFG-Parallel to a diffusion **pipeline**. We use the Qwen-Image pipeline (`vllm_omni/diffusion/models/qwen_image/pipeline_qwen_image.py`) as the reference implementation.

In `QwenImagePipeline`, each diffusion step runs two denoiser forward passes sequentially:

- positive (prompt-conditioned)
- negative (negative-prompt-conditioned)

CFG-Parallel assigns these two branches to different ranks in the **CFG group** and synchronizes the results.

vLLM-omni provides `CFGParallelMixin` base class that encapsulates the CFG parallel logic. By inheriting from this mixin and calling its methods, pipelines can easily implement CFG parallel without writing repetitive code.

**Key Methods in CFGParallelMixin:**
- `predict_noise_maybe_with_cfg()`: Automatically handles CFG parallel noise prediction
- `scheduler_step_maybe_with_cfg()`: Scheduler step with automatic CFG rank synchronization

**Example Implementation:**

```python
class QwenImageCFGParallelMixin(CFGParallelMixin):
    """
    Base Mixin class for Qwen Image pipelines providing shared CFG methods.
    """

    def diffuse(
        self,
        prompt_embeds: torch.Tensor,
        prompt_embeds_mask: torch.Tensor,
        negative_prompt_embeds: torch.Tensor,
        negative_prompt_embeds_mask: torch.Tensor,
        latents: torch.Tensor,
        img_shapes: torch.Tensor,
        txt_seq_lens: torch.Tensor,
        negative_txt_seq_lens: torch.Tensor,
        timesteps: torch.Tensor,
        do_true_cfg: bool,
        guidance: torch.Tensor,
        true_cfg_scale: float,
        image_latents: torch.Tensor | None = None,
        cfg_normalize: bool = True,
        additional_transformer_kwargs: dict[str, Any] | None = None,
    ) -> torch.Tensor:
        self.transformer.do_true_cfg = do_true_cfg

        for i, t in enumerate(timesteps):
            timestep = t.expand(latents.shape[0]).to(device=latents.device, dtype=latents.dtype)

            # Prepare kwargs for positive (conditional) prediction
            positive_kwargs = {
                "hidden_states": latents,
                "timestep": timestep / 1000,
                "guidance": guidance,
                "encoder_hidden_states_mask": prompt_embeds_mask,
                "encoder_hidden_states": prompt_embeds,
                "img_shapes": img_shapes,
                "txt_seq_lens": txt_seq_lens,
            }

            # Prepare kwargs for negative (unconditional) prediction
            if do_true_cfg:
                negative_kwargs = {
                    "hidden_states": latents,
                    "timestep": timestep / 1000,
                    "guidance": guidance,
                    "encoder_hidden_states_mask": negative_prompt_embeds_mask,
                    "encoder_hidden_states": negative_prompt_embeds,
                    "img_shapes": img_shapes,
                    "txt_seq_lens": negative_txt_seq_lens,
                }
            else:
                negative_kwargs = None

            # Predict noise with automatic CFG parallel handling
            # - In CFG parallel mode: rank0 computes positive, rank1 computes negative
            # - Automatically gathers results and combines them on rank0
            noise_pred = self.predict_noise_maybe_with_cfg(
                do_true_cfg=do_true_cfg,
                true_cfg_scale=true_cfg_scale,
                positive_kwargs=positive_kwargs,
                negative_kwargs=negative_kwargs,
                cfg_normalize=cfg_normalize,
            )

            # Step scheduler with automatic CFG synchronization
            # - Only rank0 computes the scheduler step
            # - Automatically broadcasts updated latents to all ranks
            latents = self.scheduler_step_maybe_with_cfg(
                noise_pred, t, latents, do_true_cfg
            )

        return latents
```

**How it works:**
1. Prepare separate `positive_kwargs` and `negative_kwargs` for conditional and unconditional predictions
2. Call `predict_noise_maybe_with_cfg()` which:
   - Detects if CFG parallel is enabled (`get_classifier_free_guidance_world_size() > 1`)
   - Distributes computation: rank0 processes positive, rank1 processes negative
   - Gathers predictions and combines them using `combine_cfg_noise()` on rank0
   - Returns combined noise prediction (only valid on rank0)
3. Call `scheduler_step_maybe_with_cfg()` which:
   - Only rank0 computes the scheduler step
   - Broadcasts the updated latents to all ranks for synchronization

**How to customize**

Some pipelines may need to customize the following functions in `CFGParallelMixin`:
1. You may need to edit `predict_noise` function for custom behaviors.
```python
def predict_noise(self, *args, **kwargs):
    """
    Forward pass through transformer to predict noise.

    Subclasses should override this if they need custom behavior,
    but the default implementation calls self.transformer.
    """
    return self.transformer(*args, **kwargs)[0]

```
2. The default normalization function after combining the noise predictions from both branches is as follows. You may need to customize it.
```python
def cfg_normalize_function(self, noise_pred, comb_pred):
    """
    Normalize the combined noise prediction.

    Args:
        noise_pred: positive noise prediction
        comb_pred: combined noise prediction after CFG

    Returns:
        Normalized noise prediction tensor
    """
    cond_norm = torch.norm(noise_pred, dim=-1, keepdim=True)
    noise_norm = torch.norm(comb_pred, dim=-1, keepdim=True)
    noise_pred = comb_pred * (cond_norm / noise_norm)
    return noise_pred
```