remove use_fast in AutoTokenizer (#747)

remote use_fast prarameter to adapt to paddlenlp 3.0.0.beta

---------

Co-authored-by: nifeng <nemonameless@qq.com>

Author: warrentdrew

ppdiffusers/examples/text_to_image/README_sdxl.md

@@ -119,6 +119,25 @@ export FLAGS_npu_storage_format=0
 export FLAGS_use_stride_kernel=0
 ```
 
+注意NPU训练暂不支持enable_xformers_memory_efficient_attention选项，启动命令如下:
+```bash
+python -u train_text_to_image_sdxl.py \
+  --pretrained_model_name_or_path=$MODEL_NAME \
+  --pretrained_vae_model_name_or_path=$VAE_NAME \
+  --dataset_name=$DATASET_NAME \
+  --resolution=512 --center_crop --random_flip \
+  --proportion_empty_prompts=0.2 \
+  --train_batch_size=1 \
+  --gradient_accumulation_steps=4 --gradient_checkpointing \
+  --max_train_steps=10000 \
+  --learning_rate=1e-06 --lr_scheduler="constant" --lr_warmup_steps=0 \
+  --mixed_precision="fp16" \
+  --report_to="wandb" \
+  --validation_prompt="a cute Sundar Pichai creature" --validation_epochs 5 \
+  --checkpointing_steps=5000 \
+  --output_dir="sdxl-pokemon-model"
+```
+
 
 ## Stable Diffusion XL (SDXL) LoRA 训练示例
 
@@ -200,7 +219,7 @@ python -u train_text_to_image_lora_sdxl.py \
 
 ### 推理
 
-一旦你使用上面的命令训练了一个模型，推理可以简单地使用 `StableDiffusionXLPipeline` 在加载训练好的 LoRA 权重后进行。你需要传递 `output_dir` 来加载 LoRA 权重，在这个案例中，是 `sd-pokemon-model-lora-sdxl`。
+一旦你使用上面的命令训练了一个模型，推理可以简单地使用 `StableDiffusionXLPipeline` 在加载训练好的 LoRA 权重后进行。通过修改推理脚本中的model_path变量，可以传递需要加载的 LoRA 训练权重，在这个案例中，是 `sd-pokemon-model-lora-sdxl`。
 
 ```python
 from ppdiffusers import StableDiffusionXLPipeline
@@ -224,6 +243,8 @@ import os
 
 dir_name = "your-checkpoints-path/sd-pokemon-model-lora-sdxl/"
 for file_name in sorted(os.listdir(dir_name)):
+    if 'checkpoint' not in file_name:
+        continue
     print(file_name)
     model_path = os.path.join(dir_name, file_name)
     pipe = StableDiffusionXLPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", paddle_dtype=paddle.float16)

ppdiffusers/examples/text_to_image/train_text_to_image_lora_sdxl.py

@@ -566,16 +566,10 @@ def main(args):
 
     # Load the tokenizers
     tokenizer_one = AutoTokenizer.from_pretrained(
-        args.pretrained_model_name_or_path,
-        subfolder="tokenizer",
-        revision=args.revision,
-        use_fast=False,
+        args.pretrained_model_name_or_path, subfolder="tokenizer", revision=args.revision
     )
     tokenizer_two = AutoTokenizer.from_pretrained(
-        args.pretrained_model_name_or_path,
-        subfolder="tokenizer_2",
-        revision=args.revision,
-        use_fast=False,
+        args.pretrained_model_name_or_path, subfolder="tokenizer_2", revision=args.revision
     )
 
     # import correct text encoder classes