Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-20 01:31:13 +02:00
Code Issues Packages Projects Releases Wiki Activity

Enable Triton MM on all GPUs

Browse Source
This commit is contained in:
Disty0
2026-07-14 22:21:53 +03:00
parent 1ec48189ea
commit 39d74cca18
1 changed files with 9 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/sdnq/kernel_wrappers.py
+9 -4
View File
@@ -4,7 +4,7 @@ import os
import sys
import torch
from modules import devices
from modules import devices, shared
if os.environ.get("SDNQ_ALLOW_FP8_MM", None) is None:
@@ -43,7 +43,7 @@ else:
use_openvino_mm = bool(os.environ.get("SDNQ_USE_OPENVINO_MM", "0").lower() not in {"0", "false", "no"})
if os.environ.get("SDNQ_USE_TRITON_MM", None) is None:
use_triton_mm = bool(not is_alchemist_or_igpu and (is_rdna2_and_older or devices.backend in {"zluda", "ipex", "xpu"}))
use_triton_mm = bool(not is_alchemist_or_igpu and (devices.backend in {"cuda", "rocm", "ipex", "xpu", "zluda"}))
else:
use_triton_mm = bool(os.environ.get("SDNQ_USE_TRITON_MM", "0").lower() not in {"0", "false", "no"})
@@ -102,11 +102,16 @@ elif use_triton_mm:
fp8_mm_func = sdnq_triton_mm
fp8_scaled_mm_func = sdnq_scaled_mm
use_tensorwise_fp8_matmul = True
except Exception:
except Exception as e:
use_triton_mm = False
shared.log.warning(f"SDNQ: Triton kernels are not available! Falling back to PyTorch Eager kernels. Error message: {e}")
if fp_mm_func is None and not is_alchemist_or_igpu and os.environ.get("SDNQ_USE_TRITON_MM", "1").lower() not in {"0", "false", "no"}:
if (
fp_mm_func is None and not is_alchemist_or_igpu
and devices.backend in {"cuda", "rocm", "ipex", "xpu", "zluda"}
and os.environ.get("SDNQ_USE_TRITON_MM", "1").lower() not in {"0", "false", "no"}
):
try:
from .kernels.triton_mm import sdnq_triton_mm
from .kernels.triton_scaled_mm import sdnq_scaled_mm