export MODEL_NAME="models/Diffusion_Transformer/Wan2.1-T2V-1.3B/"
export DATASET_NAME="datasets/internal_datasets/"
export DATASET_META_NAME="datasets/internal_datasets/metadata.json"
# NCCL_IB_DISABLE=1 and NCCL_P2P_DISABLE=1 are used in multi nodes without RDMA. 
# export NCCL_IB_DISABLE=1
# export NCCL_P2P_DISABLE=1
NCCL_DEBUG=INFO

# Causal-Forcing Stage 1: Autoregressive Diffusion Training (teacher forcing).
# - num_frame_per_block=3 -> chunkwise; set to 1 for the framewise variant.
# - shift=5.0 mirrors the Causal-Forcing default scheduler shift.
accelerate launch --mixed_precision="bf16" --use_fsdp \
    --fsdp_auto_wrap_policy TRANSFORMER_BASED_WRAP \
    --fsdp_transformer_layer_cls_to_wrap=CasualWanAttentionBlock \
    --fsdp_sharding_strategy "FULL_SHARD" --fsdp_state_dict_type=SHARDED_STATE_DICT \
    --fsdp_backward_prefetch "BACKWARD_PRE" --fsdp_cpu_ram_efficient_loading False \
    scripts/wan2.1_causal_forcing/train_ar_diffusion.py \
  --config_path="config/wan2.1/wan_civitai.yaml" \
  --pretrained_model_name_or_path=$MODEL_NAME \
  --train_data_dir=$DATASET_NAME \
  --train_data_meta=$DATASET_META_NAME \
  --image_sample_size=640 \
  --video_sample_size=640 \
  --token_sample_size=640 \
  --fix_sample_size 480 832 \
  --video_sample_stride=2 \
  --video_sample_n_frames=81 \
  --train_batch_size=1 \
  --gradient_accumulation_steps=1 \
  --dataloader_num_workers=8 \
  --num_train_epochs=100 \
  --checkpointing_steps=200 \
  --learning_rate=2e-06 \
  --lr_scheduler="constant_with_warmup" \
  --lr_warmup_steps=100 \
  --seed=42 \
  --output_dir="output_dir_wan2.1_causal_forcing_ar_diffusion" \
  --gradient_checkpointing \
  --mixed_precision="bf16" \
  --adam_weight_decay=3e-2 \
  --adam_epsilon=1e-10 \
  --vae_mini_batch=1 \
  --max_grad_norm=0.05 \
  --random_hw_adapt \
  --training_with_video_token_length \
  --enable_bucket \
  --num_frame_per_block=3 \
  --train_sampling_steps=1000 \
  --shift=5.0 \
  --use_timestep_weight \
  --trainable_modules "."