[AMD][Feature] support fp4 dispatch and fp8 combine in moriep (#19757)

Co-authored-by: Duyi-Wang <duyi.wang@amd.com>
This commit is contained in:
AMD-yanfeiwang
2026-03-09 12:52:05 -07:00
committed by GitHub
co-authored by Duyi-Wang
parent ffb4b6f4c1
commit f0153ad225
4 changed files with 265 additions and 23 deletions
+2
View File
@@ -75,6 +75,8 @@ SGLang supports various environment variables that can be used to configure its
| Environment Variable | Description | Default Value |
| --- | --- | --- |
| `SGLANG_MORI_FP8_DISP` | Use FP8 for dispatch | `"false"` |
| `SGLANG_MORI_FP4_DISP` | Use MXFP4 for dispatch | `"false"` |
| `SGLANG_MORI_FP8_COMB` | Use FP8 for combine | `"false"` |
| `SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK` | Maximum number of dispatch tokens per rank for MORI-EP buffer allocation | `4096` |
| `SGLANG_MORI_DISPATCH_INTER_KERNEL_SWITCH_THRESHOLD` | Threshold for switching between `InterNodeV1` and `InterNodeV1LL` kernel types. `InterNodeV1LL` is used if `SGLANG_MORI_NUM_MAX_DISPATCH_TOKENS_PER_RANK` is less than or equal to this threshold; otherwise, `InterNodeV1` is used. | `256` |
| `SGLANG_MORI_QP_PER_TRANSFER` | Number of RDMA Queue Pairs (QPs) used per transfer operation | `1` |