biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[MoE Refactor] Split up compressed_tensors_moe.py (#38960)

Browse Source 
Signed-off-by: Bill Nell <bnell@redhat.com>
This commit is contained in:
bnellnm
2026-04-06 20:07:54 -04:00
committed by GitHub
parent 00d7b497b3
commit b2b2c5239e
12 changed files with 2770 additions and 2543 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
docs/design/moe_kernel_features.md
View File

@@ -57,8 +57,8 @@ Modular kernels are supported by the following `FusedMoEMethodBase` classes.
- [`ModelOptFp8MoEMethod`][vllm.model_executor.layers.quantization.modelopt.ModelOptFp8MoEMethod]
- [`Fp8MoEMethod`][vllm.model_executor.layers.quantization.fp8.Fp8MoEMethod]
- [`CompressedTensorsW4A4Nvfp4MoEMethod`][vllm.model_executor.layers.quantization.compressed_tensors.compressed_tensors_moe.CompressedTensorsW4A4Nvfp4MoEMethod]
- [`CompressedTensorsW8A8Fp8MoEMethod`][vllm.model_executor.layers.quantization.compressed_tensors.compressed_tensors_moe.CompressedTensorsW8A8Fp8MoEMethod]
- [`CompressedTensorsW4A4Nvfp4MoEMethod`][vllm.model_executor.layers.quantization.compressed_tensors.compressed_tensors_moe.compressed_tensors_moe_w4a4_nvfp4.CompressedTensorsW4A4Nvfp4MoEMethod]
- [`CompressedTensorsW8A8Fp8MoEMethod`][vllm.model_executor.layers.quantization.compressed_tensors.compressed_tensors_moe.compressed_tensors_moe_w8a8_fp8.CompressedTensorsW8A8Fp8MoEMethod]
- [`Mxfp4MoEMethod`][vllm.model_executor.layers.quantization.mxfp4.Mxfp4MoEMethod]
- [`UnquantizedFusedMoEMethod`][vllm.model_executor.layers.fused_moe.layer.UnquantizedFusedMoEMethod]