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

cleanup flux loader

Browse Source 
Signed-off-by: Vladimir Mandic <mandic00@live.com>
This commit is contained in:
Vladimir Mandic
2025-08-11 13:38:06 -04:00
parent c92e329234
commit 87bd347116
19 changed files with 739 additions and 520 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/model_quant.py
+4
View File
@@ -642,11 +642,15 @@ def get_dit_args(load_config:dict={}, module:str=None, device_map:bool=False, al
def do_post_load_quant(sd_model, allow=True):
from modules import shared
if shared.opts.sdnq_quantize_weights and (shared.opts.sdnq_quantize_mode == 'post' or (allow and shared.opts.sdnq_quantize_mode == 'auto')):
shared.log.debug('Load model: post_quant=sdnq')
sd_model = sdnq_quantize_weights(sd_model)
if len(shared.opts.optimum_quanto_weights) > 0:
shared.log.debug('Load model: post_quant=quanto')
sd_model = optimum_quanto_weights(sd_model)
if shared.opts.torchao_quantization and (shared.opts.torchao_quantization_mode == 'post' or (allow and shared.opts.torchao_quantization_mode == 'auto')):
shared.log.debug('Load model: post_quant=torchao')
sd_model = torchao_quantization(sd_model)
if shared.opts.layerwise_quantization:
shared.log.debug('Load model: post_quant=layerwise')
apply_layerwise(sd_model)
return sd_model