Fixed API change caused issue in stable diffusion distillation for quantization example. (intel#1080)

Signed-off-by: Ye, Xinyu <xinyu.ye@intel.com>

Author: XinyuYe-Intel

examples/huggingface/pytorch/textual-inversion/distillation_for_quantization/README.md

@@ -14,7 +14,7 @@ pip install -r requirements.txt
 
 ## Get a FP32 Textual Inversion model
 
-Please refer to this [Textual Inversion example](../README.md) in Intel® Neural Compressor for more details.
+Please refer to this [Textual Inversion example](https://github.com/intel/neural-compressor/blob/master/examples/pytorch/diffusion_model/diffusers/textual_inversion/README.md) in Intel® Neural Compressor for more details.
 
 ## Do distillation for quantization
 

examples/huggingface/pytorch/textual-inversion/distillation_for_quantization/requirements.txt

@@ -1,5 +1,5 @@
 diffusers==0.4.1
-accelerate
+accelerate<=0.19.0
 torchvision
 transformers~=4.21.0
 ftfy

examples/huggingface/pytorch/textual-inversion/distillation_for_quantization/textual_inversion.py

@@ -857,9 +857,6 @@ def train_func(model):
             f"Optimized model saved to: {args.output_dir}."
         )
 
-        # change to framework model for further use
-        model = model.model
-
     # Create the pipeline using using the trained modules and save it.
     templates = imagenet_style_templates_small if args.learnable_property == "style" else imagenet_templates_small
     prompt = templates[0].format(args.placeholder_token)