Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 17:24:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

SDNQ show a warining for no Triton instead and stop overriding matmul to false

Browse Source
This commit is contained in:
Disty0
2026-06-17 02:17:21 +03:00
parent 37fc1da9ba
commit 8214b044e3
5 changed files with 4 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/model_quant.py
-10
View File
@@ -98,11 +98,6 @@ def create_sdnq_config(kwargs = None, allow: bool = True, module: str = 'Model',
from modules import shared
if allow and (shared.opts.sdnq_quantize_mode in {'pre', 'auto'}) and (module == 'any' or module in shared.opts.sdnq_quantize_weights):
from modules.sdnq import SDNQConfig
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:
log.warning('SDNQ Quantized MatMul requires a working Triton install. Disabling Quantized MatMul.')
shared.opts.sdnq_use_quantized_matmul = False
if weights_dtype is None:
if module in {"TE", "LLM"} and shared.opts.sdnq_quantize_weights_mode_te not in {"Same as model", "default"}:
@@ -354,7 +349,6 @@ def sdnq_quantize_model(model, op=None, sd_model=None, do_gc: bool = True, weigh
global quant_last_model_name, quant_last_model_device # pylint: disable=global-statement
from modules import devices, shared, timer
from modules.sdnq import sdnq_post_load_quant
from modules.sdnq.common import use_torch_compile as sdnq_use_torch_compile
if (
hasattr(model, "quantization_config")
@@ -364,10 +358,6 @@ def sdnq_quantize_model(model, op=None, sd_model=None, do_gc: bool = True, weigh
log.warning(f'Quantization: Trying to quantize a pre-quantized model. Skipping quantization of module="{op if op is not None else model.__class__}"')
return model
if shared.opts.sdnq_use_quantized_matmul and not sdnq_use_torch_compile:
log.warning('SDNQ Quantized MatMul requires a working Triton install. Disabling Quantized MatMul.')
shared.opts.sdnq_use_quantized_matmul = False
if weights_dtype is None:
if (op is not None) and ("text_encoder" in op or op in {"TE", "LLM"}) and (shared.opts.sdnq_quantize_weights_mode_te not in {"Same as model", "default"}):
weights_dtype = shared.opts.sdnq_quantize_weights_mode_te