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

Warn and disable quantized matmul if triton is not available

Browse Source
This commit is contained in:
Disty0
2025-11-29 01:34:54 +03:00
parent 3e52009a4f
commit 01a0f6b356
5 changed files with 19 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/sd_models.py
+4
View File
@@ -167,6 +167,10 @@ def set_diffuser_options(sd_model, vae=None, op:str='model', offload:bool=True,
for module_name in get_module_names(sd_model):
module = getattr(sd_model, module_name, None)
if hasattr(module, "quantization_config") and getattr(module.quantization_config, "quant_method", None) == "sdnq":
from modules.sdnq.common import use_torch_compile as sdnq_use_torch_compile
if shared.opts.sdnq_use_quantized_matmul and not sdnq_use_torch_compile:
shared.log.warning('SDNQ Quantized MatMul requires a working Triton install. Disabling Quantized MatMul.')
shared.opts.sdnq_use_quantized_matmul = False
if module.quantization_config.use_quantized_matmul != shared.opts.sdnq_use_quantized_matmul:
from modules.sdnq.loader import apply_sdnq_options_to_model
shared.log.debug(f'Setting {op} {module_name}: sdnq_use_quantized_matmul={shared.opts.sdnq_use_quantized_matmul}')