From 6e794724178423de53e0e0ad697f0d66780cefc5 Mon Sep 17 00:00:00 2001 From: OleehyO Date: Wed, 1 Jan 2025 14:53:45 +0000 Subject: [PATCH] feat: add training launch scripts for I2V and T2V models Add two shell scripts to simplify model training: - accelerate_train_i2v.sh: Launch script for Image-to-Video training - accelerate_train_t2v.sh: Launch script for Text-to-Video training Both scripts provide comprehensive configurations for: - Model settings - Data pipeline - Training parameters - System resources - Checkpointing - Validation --- finetune/accelerate_train_i2v.sh | 46 ++++++++++++++++++++++++++++++++ finetune/accelerate_train_t2v.sh | 45 +++++++++++++++++++++++++++++++ 2 files changed, 91 insertions(+) create mode 100644 finetune/accelerate_train_i2v.sh create mode 100644 finetune/accelerate_train_t2v.sh diff --git a/finetune/accelerate_train_i2v.sh b/finetune/accelerate_train_i2v.sh new file mode 100644 index 0000000..372d2c4 --- /dev/null +++ b/finetune/accelerate_train_i2v.sh @@ -0,0 +1,46 @@ +#!/usr/bin/env bash + +# 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" \ + --gen_fps 15 diff --git a/finetune/accelerate_train_t2v.sh b/finetune/accelerate_train_t2v.sh new file mode 100644 index 0000000..bdb0140 --- /dev/null +++ b/finetune/accelerate_train_t2v.sh @@ -0,0 +1,45 @@ +#!/usr/bin/env bash + +# 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" \ + --gen_fps 15