[WideEP] Remove pplx all2all backend (#33724)

Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Tyler Michael Smith
2026-02-26 14:30:10 -08:00
committed by GitHub
co-authored by Claude Opus 4.6
parent 0f2f24c8b2
commit eb19955c37
39 changed files with 107 additions and 2069 deletions
+3 -4
View File
@@ -15,7 +15,7 @@ Based on the format of the input activations, FusedMoE implementations are broad
The input activation format completely depends on the All2All Dispatch being used.
* In the Contiguous variant, the All2All Dispatch returns the activations as a contiguous tensor of shape (M, K) along with TopK Ids and TopK weights of shape (M, num_topk). Look at `DeepEPHTPrepareAndFinalize` for an example.
* In the Batched variant, the All2All Dispatch returns the activations as a tensor of shape (num_experts, max_tokens, K). Here, the activations/tokens that subscribe to the same expert are batched together. Note that not all entries of the tensor are valid. The activations tensor is typically accompanied by an `expert_num_tokens` tensor of size `num_experts`, where `expert_num_tokens[i]` indicates the number of valid tokens that subscribe to the ith expert. Look at `PplxPrepareAndFinalize` or `DeepEPLLPrepareAndFinalize` for an example.
* In the Batched variant, the All2All Dispatch returns the activations as a tensor of shape (num_experts, max_tokens, K). Here, the activations/tokens that subscribe to the same expert are batched together. Note that not all entries of the tensor are valid. The activations tensor is typically accompanied by an `expert_num_tokens` tensor of size `num_experts`, where `expert_num_tokens[i]` indicates the number of valid tokens that subscribe to the ith expert. Look at `DeepEPLLPrepareAndFinalize` for an example.
The FusedMoE operation is generally made of multiple operations, in both the Contiguous and Batched variants, as described in the diagrams below
@@ -132,7 +132,6 @@ class FusedMoEModularKernel:
Typically a FusedMoEPrepareAndFinalize type is backed by an All2All Dispatch & Combine implementation / kernel. For example,
* PplxPrepareAndFinalize type is backed by Pplx All2All kernels,
* DeepEPHTPrepareAndFinalize type is backed by DeepEP High-Throughput All2All kernels, and
* DeepEPLLPrepareAndFinalize type is backed by DeepEP Low-Latency All2All kernels.
@@ -229,7 +228,7 @@ Doing this will add the new implementation to the test suite.
### How To Check `FusedMoEPrepareAndFinalize` & `FusedMoEPermuteExpertsUnpermute` Compatibility
The unit test file [test_modular_kernel_combinations.py](../../tests/kernels/moe/test_modular_kernel_combinations.py) can also be executed as a standalone script.
Example: `python3 -m tests.kernels.moe.test_modular_kernel_combinations --pf-type PplxPrepareAndFinalize --experts-type BatchedTritonExperts`
Example: `python3 -m tests.kernels.moe.test_modular_kernel_combinations --pf-type DeepEPLLPrepareAndFinalize --experts-type BatchedTritonExperts`
As a side effect, this script can be used to test `FusedMoEPrepareAndFinalize` & `FusedMoEPermuteExpertsUnpermute` compatibility. When invoked
with incompatible types, the script will error.
@@ -238,7 +237,7 @@ with incompatible types, the script will error.
Please take a look at [profile_modular_kernel.py](../../tests/kernels/moe/modular_kernel_tools/profile_modular_kernel.py)
The script can be used to generate Torch traces for a single `FusedMoEModularKernel::forward()` call for any compatible
`FusedMoEPrepareAndFinalize` and `FusedMoEPermuteExpertsUnpermute` types.
Example: `python3 -m tests.kernels.moe.modular_kernel_tools.profile_modular_kernel --pf-type PplxPrepareAndFinalize --experts-type BatchedTritonExperts`
Example: `python3 -m tests.kernels.moe.modular_kernel_tools.profile_modular_kernel --pf-type DeepEPLLPrepareAndFinalize --experts-type BatchedTritonExperts`
## FusedMoEPrepareAndFinalize Implementations
+2 -3
View File
@@ -33,7 +33,6 @@ th {
| Backend | Output act. format | Quant. types | Quant. format | Async | Apply Weight On Input | Subclass |
|---------|--------------------|--------------|---------------|-------|-----------------------|-----------|
| naive | standard | all<sup>1</sup> | G,A,T | N | <sup>6</sup> | [layer.py][vllm.model_executor.layers.fused_moe.layer.FusedMoE] |
| pplx | batched | fp8,int8 | G,A,T | Y | Y | [`PplxPrepareAndFinalize`][vllm.model_executor.layers.fused_moe.pplx_prepare_finalize.PplxPrepareAndFinalize] |
| deepep_high_throughput | standard | fp8 | G(128),A,T<sup>2</sup> | Y | Y | [`DeepEPHTPrepareAndFinalize`][vllm.model_executor.layers.fused_moe.deepep_ht_prepare_finalize.DeepEPHTPrepareAndFinalize] |
| deepep_low_latency | batched | fp8 | G(128),A,T<sup>3</sup> | Y | Y | [`DeepEPLLPrepareAndFinalize`][vllm.model_executor.layers.fused_moe.deepep_ll_prepare_finalize.DeepEPLLPrepareAndFinalize] |
| flashinfer_all2allv | standard | nvfp4,fp8 | G,A,T | N | N | [`FlashInferA2APrepareAndFinalize`][vllm.model_executor.layers.fused_moe.flashinfer_a2a_prepare_finalize.FlashInferA2APrepareAndFinalize] |
@@ -68,7 +67,7 @@ Modular kernels are supported by the following `FusedMoEMethodBase` classes.
There are a number of MoE experts kernel implementations for different quantization types and architectures. Most follow the general API of the base Triton [`fused_experts`][vllm.model_executor.layers.fused_moe.fused_moe.fused_experts] function. Many have modular kernel adapters, so they can be used with compatible all2all backends. This table lists each experts kernel and its particular properties.
Each kernel must be provided with one of the supported input activation formats. Some flavors of kernels support both standard and batched formats through different entry points, e.g. `TritonExperts` and `BatchedTritonExperts`. Batched format kernels are currently only needed for matching with certain all2all backends, e.g. `pplx` and `DeepEPLLPrepareAndFinalize`.
Each kernel must be provided with one of the supported input activation formats. Some flavors of kernels support both standard and batched formats through different entry points, e.g. `TritonExperts` and `BatchedTritonExperts`. Batched format kernels are currently only needed for matching with certain all2all backends, e.g. `DeepEPLLPrepareAndFinalize`.
Similar to the backend kernels, each experts kernel only supports certain quantization formats. For non-modular experts, the activations will be in the original type and quantized internally by the kernel. Modular experts will expect the activations to already be in the quantized format. Both types of experts will yield outputs in the original activation type.
@@ -110,5 +109,5 @@ The following table shows "families" of modular kernels that are intended to wor
| backend | `FusedMoEPrepareAndFinalize` subclasses | `FusedMoEPermuteExpertsUnpermute` subclasses |
|---------|-----------------------------------------|----------------------------------------------|
| deepep_high_throughput | `DeepEPHTPrepareAndFinalize` | `DeepGemmExperts`,</br>`TritonExperts`,</br>`TritonOrDeepGemmExperts`,</br>`CutlassExpertsFp8`, </br>`MarlinExperts` |
| deepep_low_latency,</br>pplx | `DeepEPLLPrepareAndFinalize`,</br>`PplxPrepareAndFinalize` | `BatchedDeepGemmExperts`,</br>`BatchedTritonExperts`,</br>`CutlassBatchedExpertsFp8`,</br>`BatchedMarlinExperts` |
| deepep_low_latency | `DeepEPLLPrepareAndFinalize` | `BatchedDeepGemmExperts`,</br>`BatchedTritonExperts`,</br>`CutlassBatchedExpertsFp8`,</br>`BatchedMarlinExperts` |
| flashinfer | `FlashInferCutlassMoEPrepareAndFinalize` | `FlashInferExperts` |
+1 -1
View File
@@ -154,7 +154,7 @@ If you have PRs touching the area, please feel free to ping the area owner for r
- FlashAttention: @LucasWilkinson
- FlashInfer: @LucasWilkinson, @mgoin, @WoosukKwon
- Blackwell Kernels: @mgoin, @yewentao256
- DeepEP/DeepGEMM/pplx: @mgoin, @yewentao256
- DeepEP/DeepGEMM: @mgoin, @yewentao256
### Integrations
+3 -6
View File
@@ -8,7 +8,7 @@ EP is typically coupled with Data Parallelism (DP). While DP can be used indepen
Before using EP, you need to install the necessary dependencies. We are actively working on making this easier in the future:
1. **Install DeepEP and pplx-kernels**: Set up host environment following vLLM's guide for EP kernels [here](../../tools/ep_kernels).
1. **Install DeepEP**: Set up host environment following vLLM's guide for EP kernels [here](../../tools/ep_kernels).
2. **Install DeepGEMM library**: Follow the [official instructions](https://github.com/deepseek-ai/DeepGEMM#installation).
3. **For disaggregated serving**: Install `gdrcopy` by running the [`install_gdrcopy.sh`](../../tools/install_gdrcopy.sh) script (e.g., `install_gdrcopy.sh "${GDRCOPY_OS_VERSION}" "12.8" "x64"`). You can find available OS versions [here](https://developer.download.nvidia.com/compute/redist/gdrcopy/CUDA%2012.8/).
@@ -19,7 +19,6 @@ vLLM provides multiple communication backends for EP. Use `--all2all-backend` to
| Backend | Use Case | Features | Best For |
|---------|----------|----------|----------|
| `allgather_reducescatter` | Default backend | Standard all2all using allgather/reducescatter primitives | General purpose, works with any EP+DP configuration |
| `pplx` | Single node | Chunked prefill support, efficient intra-node communication | Single-node deployments, development |
| `deepep_high_throughput` | Multi-node prefill | Grouped GEMM with continuous layout, optimized for prefill | Prefill-dominated workloads, high-throughput scenarios |
| `deepep_low_latency` | Multi-node decode | CUDA graph support, masked layout, optimized for decode | Decode-dominated workloads, low-latency scenarios |
| `flashinfer_all2allv` | MNNVL systems | FlashInfer alltoallv kernels for multi-node NVLink | Systems with NVLink across nodes |
@@ -71,12 +70,11 @@ For example, with `TP=2, DP=4` (8 GPUs total):
The following command serves a `DeepSeek-V3-0324` model with 1-way tensor parallel, 8-way (attention) data parallel, and 8-way expert parallel. The attention weights are replicated across all GPUs, while the expert weights are split across GPUs. It will work on a H200 (or H20) node with 8 GPUs. For H100, you can try to serve a smaller model or refer to the multi-node deployment section.
```bash
# Single node EP deployment with pplx backend
# Single node EP deployment
vllm serve deepseek-ai/DeepSeek-V3-0324 \
--tensor-parallel-size 1 \ # Tensor parallelism across 1 GPU
--data-parallel-size 8 \ # Data parallelism across 8 processes
--enable-expert-parallel \ # Enable expert parallelism
--all2all-backend pplx # Use pplx communication backend
--enable-expert-parallel # Enable expert parallelism
```
## Multi-Node Deployment
@@ -197,7 +195,6 @@ vllm serve deepseek-ai/DeepSeek-V3-0324 \
--tensor-parallel-size 1 \ # Tensor parallelism
--data-parallel-size 8 \ # Data parallelism
--enable-expert-parallel \ # Enable EP
--all2all-backend pplx \ # Use pplx communication backend
--enable-eplb \ # Enable load balancer
--eplb-config '{"window_size":1000,"step_interval":3000,"num_redundant_experts":2,"log_balancedness":true}'
```