Create run_8x.sh

examples/stable-diffusion/run_8x.sh

@@ -0,0 +1,24 @@
+python ../gaudi_spawn.py --world_size 8 --use_mpi train_text_to_image_sdxl.py \
+  --pretrained_model_name_or_path stabilityai/stable-diffusion-xl-base-1.0 \
+  --pretrained_vae_model_name_or_path stabilityai/sdxl-vae \
+  --dataset_name lambdalabs/pokemon-blip-captions \
+  --resolution 1024 \
+  --center_crop \
+  --random_flip \
+  --proportion_empty_prompts=0.2 \
+  --train_batch_size 16 \
+  --max_train_steps 313 \
+  --learning_rate 1e-05 \
+  --max_grad_norm 1 \
+  --lr_scheduler constant \
+  --lr_warmup_steps 0 \
+  --output_dir sdxl-pokemon-model \
+  --gaudi_config_name Habana/stable-diffusion \
+  --throughput_warmup_steps 3 \
+  --dataloader_num_workers 8 \
+  --bf16 \
+  --use_hpu_graphs_for_inference \
+  --validation_prompt="a robotic cat with wings" \
+  --validation_epochs 6 \
+  --checkpointing_steps 313 \
+  --cache_dir /root/software/data/pytorch/huggingface/sdxl 2>&1 | tee log_8x_bs16.txt

examples/stable-diffusion/train_text_to_image_sdxl.py

@@ -944,6 +944,7 @@ def collate_fn(examples):
         num_training_steps=args.max_train_steps * args.gradient_accumulation_steps,
     )
 
+    unet.to("hpu")
     # Prepare everything with our `accelerator`.
     unet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
         unet, optimizer, train_dataloader, lr_scheduler