#!/bin/bash # run # accelerate config # check with # accelerate env export MODEL_DIR="PixArt-alpha/PixArt-XL-2-512x512" export OUTPUT_DIR="output/pixart-controlnet-hf-diffusers-test" accelerate launch ./train_pixart_controlnet_hf.py --mixed_precision="fp16" \ --pretrained_model_name_or_path=$MODEL_DIR \ --output_dir=$OUTPUT_DIR \ --dataset_name=fusing/fill50k \ --resolution=512 \ --learning_rate=1e-5 \ --train_batch_size=1 \ --gradient_accumulation_steps=4 \ --report_to="wandb" \ --seed=42 \ --dataloader_num_workers=8 # --lr_scheduler="cosine" --lr_warmup_steps=0 \