Reorganize training script arguments

finetune/accelerate_train_i2v.sh

@@ -3,44 +3,67 @@
 # Prevent tokenizer parallelism issues
 export TOKENIZERS_PARALLELISM=false
 
-# Launch training with accelerate
-accelerate launch train.py \
-    ########## Model Configuration ##########
-    --model_path "THUDM/CogVideoX1.5-5B-I2V" \
-    --model_name "cogvideox1.5-i2v" \
-    --model_type "i2v" \
-    --training_type "lora" \
-    
-    ########## Output Configuration ##########
-    --output_dir "/path/to/output/dir" \
-    --report_to "tensorboard" \
-    
-    ########## Data Configuration ##########
-    --data_root "/path/to/data/dir" \
-    --caption_column "prompt.txt" \
-    --video_column "videos.txt" \
-    --image_column "images.txt" \
-    --train_resolution "48x768x1360" \
-    
-    ########## Training Configuration ##########
-    --train_epochs 10 \
-    --batch_size 1 \
-    --gradient_accumulation_steps 1 \
-    --mixed_precision "bf16" \
-    --seed 42 \
-    
-    ########## System Configuration ##########
-    --num_workers 8 \
-    --pin_memory True \
-    --nccl_timeout 1800 \
-    
-    ########## Checkpointing Configuration ##########
-    --checkpointing_steps 200 \
-    --checkpointing_limit 10 \
-    
-    ########## Validation Configuration ##########
-    --do_validation False \
-    --validation_dir "path/to/validation/dir" \
-    --validation_steps 400 \
-    --validation_prompts "prompts.txt" \
+# Model Configuration
+MODEL_ARGS=(
+    --model_path "THUDM/CogVideoX1.5-5B-I2V"
+    --model_name "cogvideox1.5-i2v"
+    --model_type "i2v"
+    --training_type "lora"
+)
+
+# Output Configuration
+OUTPUT_ARGS=(
+    --output_dir "/path/to/output/dir"
+    --report_to "tensorboard"
+)
+
+# Data Configuration
+DATA_ARGS=(
+    --data_root "/path/to/data/dir"
+    --caption_column "prompt.txt"
+    --video_column "videos.txt"
+    --image_column "images.txt"
+    --train_resolution "80x768x1360"
+)
+
+# Training Configuration
+TRAIN_ARGS=(
+    --train_epochs 10
+    --batch_size 1
+    --gradient_accumulation_steps 1
+    --mixed_precision "bf16"
+    --seed 42
+)
+
+# System Configuration
+SYSTEM_ARGS=(
+    --num_workers 8
+    --pin_memory True
+    --nccl_timeout 1800
+)
+
+# Checkpointing Configuration
+CHECKPOINT_ARGS=(
+    --checkpointing_steps 200
+    --checkpointing_limit 10
+)
+
+# Validation Configuration
+VALIDATION_ARGS=(
+    --do_validation False
+    --validation_dir "/path/to/validation/dir"
+    --validation_steps 400
+    --validation_prompts "prompts.txt"
+    --validation_images "images.txt"
     --gen_fps 15
+)
+
+# Combine all arguments and launch training
+accelerate launch train.py \
+    "${MODEL_ARGS[@]}" \
+    "${OUTPUT_ARGS[@]}" \
+    "${DATA_ARGS[@]}" \
+    "${TRAIN_ARGS[@]}" \
+    "${SYSTEM_ARGS[@]}" \
+    "${CHECKPOINT_ARGS[@]}" \
+    "${VALIDATION_ARGS[@]}"

finetune/accelerate_train_t2v.sh

@@ -3,43 +3,65 @@
 # Prevent tokenizer parallelism issues
 export TOKENIZERS_PARALLELISM=false
 
-# Launch training with accelerate
-accelerate launch train.py \
-    ########## Model Configuration ##########
-    --model_path "THUDM/CogVideoX1.5-5B" \
-    --model_name "cogvideox1.5-t2v" \
-    --model_type "t2v" \
-    --training_type "lora" \
-    
-    ########## Output Configuration ##########
-    --output_dir "/path/to/output/dir" \
-    --report_to "tensorboard" \
-    
-    ########## Data Configuration ##########
-    --data_root "/path/to/data/dir" \
-    --caption_column "prompt.txt" \
-    --video_column "videos.txt" \
-    --train_resolution "48x768x1360" \
-    
-    ########## Training Configuration ##########
-    --train_epochs 10 \
-    --batch_size 1 \
-    --gradient_accumulation_steps 1 \
-    --mixed_precision "bf16" \
-    --seed 42 \
-    
-    ########## System Configuration ##########
-    --num_workers 8 \
-    --pin_memory True \
-    --nccl_timeout 1800 \
-    
-    ########## Checkpointing Configuration ##########
-    --checkpointing_steps 200 \
-    --checkpointing_limit 10 \
-    
-    ########## Validation Configuration ##########
-    --do_validation False \
-    --validation_dir "path/to/validation/dir" \
-    --validation_steps 400 \
-    --validation_prompts "prompts.txt" \
+# Model Configuration
+MODEL_ARGS=(
+    --model_path "THUDM/CogVideoX1.5-5B"
+    --model_name "cogvideox1.5-t2v"
+    --model_type "t2v"
+    --training_type "lora"
+)
+
+# Output Configuration
+OUTPUT_ARGS=(
+    --output_dir "/path/to/output/dir"
+    --report_to "tensorboard"
+)
+
+# Data Configuration
+DATA_ARGS=(
+    --data_root "/path/to/data/dir"
+    --caption_column "prompt.txt"
+    --video_column "videos.txt"
+    --train_resolution "80x768x1360"
+)
+
+# Training Configuration
+TRAIN_ARGS=(
+    --train_epochs 10
+    --batch_size 1
+    --gradient_accumulation_steps 1
+    --mixed_precision "bf16"
+    --seed 42
+)
+
+# System Configuration
+SYSTEM_ARGS=(
+    --num_workers 8
+    --pin_memory True
+    --nccl_timeout 1800
+)
+
+# Checkpointing Configuration
+CHECKPOINT_ARGS=(
+    --checkpointing_steps 200
+    --checkpointing_limit 10
+)
+
+# Validation Configuration
+VALIDATION_ARGS=(
+    --do_validation False
+    --validation_dir "/path/to/validation/dir"
+    --validation_steps 400
+    --validation_prompts "prompts.txt"
     --gen_fps 15
+)
+
+# Combine all arguments and launch training
+accelerate launch train.py \
+    "${MODEL_ARGS[@]}" \
+    "${OUTPUT_ARGS[@]}" \
+    "${DATA_ARGS[@]}" \
+    "${TRAIN_ARGS[@]}" \
+    "${SYSTEM_ARGS[@]}" \
+    "${CHECKPOINT_ARGS[@]}" \
+    "${VALIDATION_ARGS[@]}"