Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 01:04:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

Move quant functions to model_quant.py

Browse Source
This commit is contained in:
Disty0
2025-01-23 21:50:26 +03:00
parent 5a7c1f50c1
commit 9b579bfd96
5 changed files with 318 additions and 292 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/model_te.py
+2 -2
View File
@@ -64,12 +64,12 @@ def load_t5(name=None, cache_dir=None):
t5 = transformers.T5EncoderModel.from_pretrained(repo_id, subfolder='text_encoder_3', quantization_config=quantization_config, cache_dir=cache_dir, torch_dtype=devices.dtype)
elif 'qint8' in name.lower():
model_quant.load_quanto('Load model: type=T5')
from modules.sd_models_compile import optimum_quanto_model
from modules.model_quant import optimum_quanto_model
t5 = transformers.T5EncoderModel.from_pretrained(repo_id, subfolder='text_encoder_3', cache_dir=cache_dir, torch_dtype=devices.dtype)
t5 = optimum_quanto_model(t5, weights="qint8", activations="none")
elif 'int8' in name.lower():
install('nncf==2.7.0', quiet=True)
from modules.sd_models_compile import nncf_compress_model
from modules.model_quant import nncf_compress_model
from modules.sd_hijack import NNCF_T5DenseGatedActDense
t5 = transformers.T5EncoderModel.from_pretrained(repo_id, subfolder='text_encoder_3', cache_dir=cache_dir, torch_dtype=devices.dtype)
for i in range(len(t5.encoder.block)):