Upload 3 files

checkpoint-2000.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:508c34acc05a5f13e0748286b860d3bf14629f8e073c908e18c44fe2a2360449
+size 1079730682

train_lora.sh

@@ -0,0 +1,82 @@
+# export MODEL_NAME="models/Diffusion_Transformer/Wan2.1-T2V-14B"
+# 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
+
+# accelerate launch --mixed_precision="bf16" scripts/wan2.1/train_lora.py \
+#   --config_path="config/wan2.1/wan_civitai.yaml" \
+#   --pretrained_model_name_or_path=$MODEL_NAME \
+#   --train_data_dir=$DATASET_NAME \
+#   --train_data_meta=$DATASET_META_NAME \
+#   --image_sample_size=1024 \
+#   --video_sample_size=256 \
+#   --token_sample_size=512 \
+#   --video_sample_stride=2 \
+#   --video_sample_n_frames=81 \
+#   --train_batch_size=1 \
+#   --video_repeat=1 \
+#   --gradient_accumulation_steps=1 \
+#   --dataloader_num_workers=8 \
+#   --num_train_epochs=100 \
+#   --checkpointing_steps=50 \
+#   --learning_rate=1e-04 \
+#   --seed=42 \
+#   --output_dir="output_dir" \
+#   --gradient_checkpointing \
+#   --mixed_precision="bf16" \
+#   --adam_weight_decay=3e-2 \
+#   --adam_epsilon=1e-10 \
+#   --vae_mini_batch=1 \
+#   --max_grad_norm=0.05 \
+#   --random_hw_adapt \
+#   --training_with_video_token_length \
+#   --enable_bucket \
+#   --uniform_sampling \
+#   --low_vram 
+
+# Training command for I2V
+export MODEL_NAME="/mnt/data/public_ckpt/videogen_public/Wan-AI/Wan2.1-I2V-14B-480P"
+export DATASET_NAME=""
+export DATASET_META_NAME="/mnt/data/2002-ssd/filtered_clips_from_static/lora_train_combined_datasets/combined_captions_1021.json"
+# export CHECKPOINT="checkpoint-800"
+# export NCCL_IB_DISABLE=1
+# export NCCL_P2P_DISABLE=1
+NCCL_DEBUG=INFO
+
+accelerate launch --use_deepspeed --deepspeed_config_file config/zero_stage2_config.json --deepspeed_multinode_launcher standard scripts/wan2.1/train_lora.py \
+  --config_path="config/wan2.1/wan_civitai.yaml" \
+  --pretrained_model_name_or_path=$MODEL_NAME \
+  --train_data_dir=$DATASET_NAME \
+  --train_data_meta=$DATASET_META_NAME \
+  --image_sample_size=512 \
+  --video_sample_size=632 \
+  --token_sample_size=632 \
+  --video_sample_stride=1 \
+  --video_sample_n_frames=81 \
+  --train_batch_size=1 \
+  --video_repeat=1 \
+  --gradient_accumulation_steps=1 \
+  --dataloader_num_workers=8 \
+  --num_train_epochs=1000 \
+  --checkpointing_steps=500 \
+  --learning_rate=2e-05 \
+  --seed=42 \
+  --output_dir="output_dir_14b_1022_7directions_8batchsize_2e-5lr" \
+  --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 \
+  --low_vram \
+  --train_mode="i2v" \
+  --save_state \
+  # --resume_from_checkpoint="latest" \
+    # --uniform_sampling \
+