experiment 2: train from pre-trained arc2face encoder

examples/text_to_image/README.md

@@ -76,6 +76,25 @@ accelerate launch --mixed_precision="fp16"  train_text_to_image.py \
 <!-- accelerate_snippet_end -->
 
 
+accelerate launch --mixed_precision="bf16"  train_text_to_image.py \
+  --pretrained_model_name_or_path=SG161222/Realistic_Vision_V6.0_B1_noVAE \
+  --dataset_name=nielsr/CelebA-faces \
+  --resolution=512 --center_crop --random_flip \
+  --train_batch_size=64 \
+  --gradient_checkpointing \
+  --max_train_steps=60000 \
+  --learning_rate=1e-05 \
+  --max_grad_norm=1 \
+  --lr_scheduler="constant" --lr_warmup_steps=0 \
+  --output_dir="arc2face-attempt-1" \
+  --validation_steps=100 \
+  --validation_prompt="taylor.jpg" \
+  --report_to="wandb" \
+  --resume_from_checkpoint="latest" \
+  --checkpoints_total_limit=5 \
+  --checkpointing_steps=1000
+
+
 To run on your own training files prepare the dataset according to the format required by `datasets`, you can find the instructions for how to do that in this [document](https://huggingface.co/docs/datasets/v2.4.0/en/image_load#imagefolder-with-metadata).
 If you wish to use custom loading logic, you should modify the script, we have left pointers for that in the training script.