minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

[AMD][Quantization] Online MXFP4 quantization 4/N - NVFP4 to MXFP4 Online Requantization on AMD GPUs (#29328)

Browse Source
This commit is contained in:
Colin Z
2026-08-14 21:59:39 -07:00
committed by GitHub
parent 5afdb1caea
commit bc7e3ba66c
14 changed files with 1218 additions and 204 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/server_args.py
+1 -1
View File
@@ -167,7 +167,7 @@ QUANTIZATION_CHOICES = [
"mxfp_w4a8", # for NPU W4A8 (MXFP4 weights + MXFP8 activations)
"quark", # AMD Quark quantizer (FP8 / MXFP4 / Int4FP8 etc.)
"quark_int4fp8_moe",
"quark_mxfp4", # Online MOE + linear quantization.
"quark_mxfp4", # Online MOE + linear quantization (incl. NVFP4 -> MXFP4 requantization).
# Apple Silicon MLX backend — on-the-fly quantization of fp16 weights at load
# time via mlx.nn.quantize. Only takes effect when SGLANG_USE_MLX=1.
"mlx_q4", # 4 bits, group_size=64 (mlx-community default)