mirror of
https://github.com/modelscope/DiffSynth-Studio.git
synced 2026-04-24 15:06:17 +00:00
ace-step train
This commit is contained in:
mi804
18
examples/ace_step/model_training/full/Ace-Step1.5.sh
Normal file
18
examples/ace_step/model_training/full/Ace-Step1.5.sh
Normal file
@@ -0,0 +1,18 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/Ace-Step1.5/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/Ace-Step1.5/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-5 \
|
||||
--num_epochs 2 \
|
||||
--trainable_models "dit" \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/Ace-Step1.5" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/Ace-Step1.5/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/Ace-Step1.5:acestep-v15-turbo/model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/Ace-Step1.5_full" \
|
||||
--data_file_keys "audio"
|
||||
18
examples/ace_step/model_training/full/acestep-v15-base.sh
Normal file
18
examples/ace_step/model_training/full/acestep-v15-base.sh
Normal file
@@ -0,0 +1,18 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-base/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-base/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-5 \
|
||||
--num_epochs 2 \
|
||||
--trainable_models "dit" \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-base" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-base/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-base:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-base_full" \
|
||||
--data_file_keys "audio"
|
||||
18
examples/ace_step/model_training/full/acestep-v15-sft.sh
Normal file
18
examples/ace_step/model_training/full/acestep-v15-sft.sh
Normal file
@@ -0,0 +1,18 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-sft/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-sft/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-5 \
|
||||
--num_epochs 2 \
|
||||
--trainable_models "dit" \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-sft" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-sft/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-sft:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-sft_full" \
|
||||
--data_file_keys "audio"
|
||||
@@ -0,0 +1,18 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-continuous/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-turbo-continuous/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-5 \
|
||||
--num_epochs 2 \
|
||||
--trainable_models "dit" \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-continuous" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-continuous/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-turbo-continuous:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-turbo-continuous_full" \
|
||||
--data_file_keys "audio"
|
||||
@@ -0,0 +1,18 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift1/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-turbo-shift1/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-5 \
|
||||
--num_epochs 2 \
|
||||
--trainable_models "dit" \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift1" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift1/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-turbo-shift1:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-turbo-shift1_full" \
|
||||
--data_file_keys "audio"
|
||||
@@ -0,0 +1,18 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift3/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-turbo-shift3/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-5 \
|
||||
--num_epochs 2 \
|
||||
--trainable_models "dit" \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift3" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift3/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-turbo-shift3:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-turbo-shift3_full" \
|
||||
--data_file_keys "audio"
|
||||
18
examples/ace_step/model_training/full/acestep-v15-xl-base.sh
Normal file
18
examples/ace_step/model_training/full/acestep-v15-xl-base.sh
Normal file
@@ -0,0 +1,18 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-xl-base/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-xl-base/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-5 \
|
||||
--num_epochs 2 \
|
||||
--trainable_models "dit" \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-base" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-base/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-xl-base:model-*.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-xl-base_full" \
|
||||
--data_file_keys "audio"
|
||||
18
examples/ace_step/model_training/full/acestep-v15-xl-sft.sh
Normal file
18
examples/ace_step/model_training/full/acestep-v15-xl-sft.sh
Normal file
@@ -0,0 +1,18 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-xl-sft/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-xl-sft/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-5 \
|
||||
--num_epochs 2 \
|
||||
--trainable_models "dit" \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-sft" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-sft/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-xl-sft:model-*.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-xl-sft_full" \
|
||||
--data_file_keys "audio"
|
||||
@@ -0,0 +1,18 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-xl-turbo/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-xl-turbo/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-5 \
|
||||
--num_epochs 2 \
|
||||
--trainable_models "dit" \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-turbo" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-turbo/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-xl-turbo:model-*.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-xl-turbo_full" \
|
||||
--data_file_keys "audio"
|
||||
20
examples/ace_step/model_training/lora/Ace-Step1.5.sh
Normal file
20
examples/ace_step/model_training/lora/Ace-Step1.5.sh
Normal file
@@ -0,0 +1,20 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/Ace-Step1.5/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/Ace-Step1.5/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-4 \
|
||||
--num_epochs 20 \
|
||||
--lora_rank 32 \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/Ace-Step1.5" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/Ace-Step1.5/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/Ace-Step1.5:acestep-v15-turbo/model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--lora_base_model "dit" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/Ace-Step1.5_lora" \
|
||||
--lora_target_modules "q_proj,k_proj,v_proj,o_proj" \
|
||||
--data_file_keys "audio"
|
||||
20
examples/ace_step/model_training/lora/acestep-v15-base.sh
Normal file
20
examples/ace_step/model_training/lora/acestep-v15-base.sh
Normal file
@@ -0,0 +1,20 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-base/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-base/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-4 \
|
||||
--num_epochs 20 \
|
||||
--lora_rank 32 \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-base" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-base/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-base:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--lora_base_model "dit" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-base_lora" \
|
||||
--lora_target_modules "q_proj,k_proj,v_proj,o_proj" \
|
||||
--data_file_keys "audio"
|
||||
20
examples/ace_step/model_training/lora/acestep-v15-sft.sh
Normal file
20
examples/ace_step/model_training/lora/acestep-v15-sft.sh
Normal file
@@ -0,0 +1,20 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-sft/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-sft/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-4 \
|
||||
--num_epochs 20 \
|
||||
--lora_rank 32 \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-sft" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-sft/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-sft:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--lora_base_model "dit" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-sft_lora" \
|
||||
--lora_target_modules "q_proj,k_proj,v_proj,o_proj" \
|
||||
--data_file_keys "audio"
|
||||
@@ -0,0 +1,20 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-continuous/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-turbo-continuous/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-4 \
|
||||
--num_epochs 20 \
|
||||
--lora_rank 32 \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-continuous" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-continuous/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-turbo-continuous:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--lora_base_model "dit" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-turbo-continuous_lora" \
|
||||
--lora_target_modules "q_proj,k_proj,v_proj,o_proj" \
|
||||
--data_file_keys "audio"
|
||||
@@ -0,0 +1,20 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift1/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-turbo-shift1/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-4 \
|
||||
--num_epochs 20 \
|
||||
--lora_rank 32 \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift1" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift1/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-turbo-shift1:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--lora_base_model "dit" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-turbo-shift1_lora" \
|
||||
--lora_target_modules "q_proj,k_proj,v_proj,o_proj" \
|
||||
--data_file_keys "audio"
|
||||
@@ -0,0 +1,20 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift3/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-turbo-shift3/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-4 \
|
||||
--num_epochs 20 \
|
||||
--lora_rank 32 \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift3" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-turbo-shift3/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-turbo-shift3:model.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--lora_base_model "dit" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-turbo-shift3_lora" \
|
||||
--lora_target_modules "q_proj,k_proj,v_proj,o_proj" \
|
||||
--data_file_keys "audio"
|
||||
20
examples/ace_step/model_training/lora/acestep-v15-xl-base.sh
Normal file
20
examples/ace_step/model_training/lora/acestep-v15-xl-base.sh
Normal file
@@ -0,0 +1,20 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-xl-base/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-xl-base/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-4 \
|
||||
--num_epochs 20 \
|
||||
--lora_rank 32 \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-base" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-base/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-xl-base:model-*.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--lora_base_model "dit" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-xl-base_lora" \
|
||||
--lora_target_modules "q_proj,k_proj,v_proj,o_proj" \
|
||||
--data_file_keys "audio"
|
||||
20
examples/ace_step/model_training/lora/acestep-v15-xl-sft.sh
Normal file
20
examples/ace_step/model_training/lora/acestep-v15-xl-sft.sh
Normal file
@@ -0,0 +1,20 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-xl-sft/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-xl-sft/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-4 \
|
||||
--num_epochs 20 \
|
||||
--lora_rank 32 \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-sft" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-sft/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-xl-sft:model-*.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--lora_base_model "dit" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-xl-sft_lora" \
|
||||
--lora_target_modules "q_proj,k_proj,v_proj,o_proj" \
|
||||
--data_file_keys "audio"
|
||||
@@ -0,0 +1,20 @@
|
||||
# Dataset: data/diffsynth_example_dataset/ace_step/acestep-v15-xl-turbo/
|
||||
# Download: modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "ace_step/acestep-v15-xl-turbo/*" --local_dir ./data/diffsynth_example_dataset
|
||||
|
||||
accelerate launch examples/ace_step/model_training/train.py \
|
||||
--learning_rate 1e-4 \
|
||||
--num_epochs 20 \
|
||||
--lora_rank 32 \
|
||||
--use_gradient_checkpointing \
|
||||
--find_unused_parameters \
|
||||
--dataset_base_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-turbo" \
|
||||
--dataset_metadata_path "./data/diffsynth_example_dataset/ace_step/acestep-v15-xl-turbo/metadata.json" \
|
||||
--model_id_with_origin_paths "ACE-Step/acestep-v15-xl-turbo:model-*.safetensors,ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/model.safetensors,ACE-Step/Ace-Step1.5:vae/diffusion_pytorch_model.safetensors" \
|
||||
--tokenizer_path "ACE-Step/Ace-Step1.5:Qwen3-Embedding-0.6B/" \
|
||||
--silence_latent_path "ACE-Step/Ace-Step1.5:acestep-v15-turbo/silence_latent.pt" \
|
||||
--lora_base_model "dit" \
|
||||
--remove_prefix_in_ckpt "pipe.dit." \
|
||||
--dataset_repeat 50 \
|
||||
--output_path "./models/train/acestep-v15-xl-turbo_lora" \
|
||||
--lora_target_modules "q_proj,k_proj,v_proj,o_proj" \
|
||||
--data_file_keys "audio"
|
||||
173
examples/ace_step/model_training/train.py
Normal file
173
examples/ace_step/model_training/train.py
Normal file
@@ -0,0 +1,173 @@
|
||||
import torch, os, argparse, accelerate, warnings, torchaudio
|
||||
import math
|
||||
from diffsynth.core import UnifiedDataset
|
||||
from diffsynth.core.data.operators import ToAbsolutePath, RouteByType, DataProcessingOperator, LoadPureAudioWithTorchaudio
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.diffusion import *
|
||||
os.environ["TOKENIZERS_PARALLELISM"] = "false"
|
||||
|
||||
|
||||
class LoadAceStepAudio(DataProcessingOperator):
|
||||
"""Load audio file and return waveform tensor [2, T] at 48kHz."""
|
||||
def __init__(self, target_sr=48000):
|
||||
self.target_sr = target_sr
|
||||
|
||||
def __call__(self, data: str):
|
||||
try:
|
||||
waveform, sample_rate = torchaudio.load(data)
|
||||
if sample_rate != self.target_sr:
|
||||
resampler = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=self.target_sr)
|
||||
waveform = resampler(waveform)
|
||||
if waveform.shape[0] == 1:
|
||||
waveform = waveform.repeat(2, 1)
|
||||
return waveform
|
||||
except Exception as e:
|
||||
warnings.warn(f"Cannot load audio from {data}: {e}")
|
||||
return None
|
||||
|
||||
|
||||
class AceStepTrainingModule(DiffusionTrainingModule):
|
||||
def __init__(
|
||||
self,
|
||||
model_paths=None, model_id_with_origin_paths=None,
|
||||
tokenizer_path=None, silence_latent_path=None,
|
||||
trainable_models=None,
|
||||
lora_base_model=None, lora_target_modules="", lora_rank=32, lora_checkpoint=None,
|
||||
preset_lora_path=None, preset_lora_model=None,
|
||||
use_gradient_checkpointing=True,
|
||||
use_gradient_checkpointing_offload=False,
|
||||
extra_inputs=None,
|
||||
fp8_models=None,
|
||||
offload_models=None,
|
||||
device="cpu",
|
||||
task="sft",
|
||||
):
|
||||
super().__init__()
|
||||
# ===== 解析模型配置（固定写法） =====
|
||||
model_configs = self.parse_model_configs(model_paths, model_id_with_origin_paths, fp8_models=fp8_models, offload_models=offload_models, device=device)
|
||||
# ===== Tokenizer 配置 =====
|
||||
text_tokenizer_config = self.parse_path_or_model_id(tokenizer_path, default_value=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"))
|
||||
silence_latent_config = self.parse_path_or_model_id(silence_latent_path, default_value=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"))
|
||||
# ===== 构建 Pipeline =====
|
||||
self.pipe = AceStepPipeline.from_pretrained(torch_dtype=torch.bfloat16, device=device, model_configs=model_configs, text_tokenizer_config=text_tokenizer_config, silence_latent_config=silence_latent_config)
|
||||
# ===== 拆分 Pipeline Units（固定写法） =====
|
||||
self.pipe = self.split_pipeline_units(task, self.pipe, trainable_models, lora_base_model)
|
||||
|
||||
# ===== 切换到训练模式（固定写法） =====
|
||||
self.switch_pipe_to_training_mode(
|
||||
self.pipe, trainable_models,
|
||||
lora_base_model, lora_target_modules, lora_rank, lora_checkpoint,
|
||||
preset_lora_path, preset_lora_model,
|
||||
task=task,
|
||||
)
|
||||
|
||||
# ===== 其他配置（固定写法） =====
|
||||
self.use_gradient_checkpointing = use_gradient_checkpointing
|
||||
self.use_gradient_checkpointing_offload = use_gradient_checkpointing_offload
|
||||
self.extra_inputs = extra_inputs.split(",") if extra_inputs is not None else []
|
||||
self.fp8_models = fp8_models
|
||||
self.task = task
|
||||
# ===== 任务模式路由（固定写法） =====
|
||||
self.task_to_loss = {
|
||||
"sft:data_process": lambda pipe, *args: args,
|
||||
"sft": lambda pipe, inputs_shared, inputs_posi, inputs_nega: FlowMatchSFTLoss(pipe, **inputs_shared, **inputs_posi),
|
||||
"sft:train": lambda pipe, inputs_shared, inputs_posi, inputs_nega: FlowMatchSFTLoss(pipe, **inputs_shared, **inputs_posi),
|
||||
}
|
||||
|
||||
def get_pipeline_inputs(self, data):
|
||||
inputs_posi = {"prompt": data["prompt"], "positive": True}
|
||||
inputs_nega = {"positive": False}
|
||||
duration = math.floor(data['audio'][0].shape[1] / data['audio'][1]) if data.get("audio") is not None else data.get("duration", 60)
|
||||
# ===== 共享参数 =====
|
||||
inputs_shared = {
|
||||
# ===== 核心字段映射 =====
|
||||
"input_audio": data["audio"],
|
||||
# ===== 音频生成任务所需元数据 =====
|
||||
"lyrics": data["lyrics"],
|
||||
"task_type": "text2music",
|
||||
"duration": duration,
|
||||
"bpm": data.get("bpm", 100),
|
||||
"keyscale": data.get("keyscale", "C major"),
|
||||
"timesignature": data.get("timesignature", "4"),
|
||||
"vocal_language": data.get("vocal_language", "unknown"),
|
||||
# ===== 框架控制参数（固定写法） =====
|
||||
"cfg_scale": 1,
|
||||
"rand_device": self.pipe.device,
|
||||
"use_gradient_checkpointing": self.use_gradient_checkpointing,
|
||||
"use_gradient_checkpointing_offload": self.use_gradient_checkpointing_offload,
|
||||
}
|
||||
# ===== 额外字段注入：通过 --extra_inputs 配置的数据集列名（固定写法） =====
|
||||
inputs_shared = self.parse_extra_inputs(data, self.extra_inputs, inputs_shared)
|
||||
return inputs_shared, inputs_posi, inputs_nega
|
||||
|
||||
def forward(self, data, inputs=None):
|
||||
# ===== 标准实现，不要修改（固定写法） =====
|
||||
if inputs is None: inputs = self.get_pipeline_inputs(data)
|
||||
inputs = self.transfer_data_to_device(inputs, self.pipe.device, self.pipe.torch_dtype)
|
||||
for unit in self.pipe.units:
|
||||
inputs = self.pipe.unit_runner(unit, self.pipe, *inputs)
|
||||
loss = self.task_to_loss[self.task](self.pipe, *inputs)
|
||||
return loss
|
||||
|
||||
|
||||
def ace_step_parser():
|
||||
parser = argparse.ArgumentParser(description="ACE-Step training.")
|
||||
parser = add_general_config(parser)
|
||||
parser.add_argument("--tokenizer_path", type=str, default=None, help="Tokenizer path in format model_id:origin_pattern.")
|
||||
parser.add_argument("--silence_latent_path", type=str, default=None, help="Silence latent path in format model_id:origin_pattern.")
|
||||
parser.add_argument("--initialize_model_on_cpu", default=False, action="store_true", help="Whether to initialize models on CPU.")
|
||||
return parser
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
parser = ace_step_parser()
|
||||
args = parser.parse_args()
|
||||
# ===== Accelerator 配置（固定写法） =====
|
||||
accelerator = accelerate.Accelerator(
|
||||
gradient_accumulation_steps=args.gradient_accumulation_steps,
|
||||
kwargs_handlers=[accelerate.DistributedDataParallelKwargs(find_unused_parameters=args.find_unused_parameters)],
|
||||
)
|
||||
# ===== 数据集定义 =====
|
||||
dataset = UnifiedDataset(
|
||||
base_path=args.dataset_base_path,
|
||||
metadata_path=args.dataset_metadata_path,
|
||||
repeat=args.dataset_repeat,
|
||||
data_file_keys=args.data_file_keys.split(","),
|
||||
main_data_operator=None,
|
||||
special_operator_map={
|
||||
"audio": ToAbsolutePath(args.dataset_base_path) >> LoadPureAudioWithTorchaudio(target_sample_rate=48000),
|
||||
},
|
||||
)
|
||||
# ===== TrainingModule =====
|
||||
model = AceStepTrainingModule(
|
||||
model_paths=args.model_paths,
|
||||
model_id_with_origin_paths=args.model_id_with_origin_paths,
|
||||
tokenizer_path=args.tokenizer_path,
|
||||
silence_latent_path=args.silence_latent_path,
|
||||
trainable_models=args.trainable_models,
|
||||
lora_base_model=args.lora_base_model,
|
||||
lora_target_modules=args.lora_target_modules,
|
||||
lora_rank=args.lora_rank,
|
||||
lora_checkpoint=args.lora_checkpoint,
|
||||
preset_lora_path=args.preset_lora_path,
|
||||
preset_lora_model=args.preset_lora_model,
|
||||
use_gradient_checkpointing=args.use_gradient_checkpointing,
|
||||
use_gradient_checkpointing_offload=args.use_gradient_checkpointing_offload,
|
||||
extra_inputs=args.extra_inputs,
|
||||
fp8_models=args.fp8_models,
|
||||
offload_models=args.offload_models,
|
||||
task=args.task,
|
||||
device="cpu" if args.initialize_model_on_cpu else accelerator.device,
|
||||
)
|
||||
# ===== ModelLogger（固定写法） =====
|
||||
model_logger = ModelLogger(
|
||||
args.output_path,
|
||||
remove_prefix_in_ckpt=args.remove_prefix_in_ckpt,
|
||||
)
|
||||
# ===== 任务路由（固定写法） =====
|
||||
launcher_map = {
|
||||
"sft:data_process": launch_data_process_task,
|
||||
"sft": launch_training_task,
|
||||
"sft:train": launch_training_task,
|
||||
}
|
||||
launcher_map[args.task](accelerator, dataset, model, model_logger, args=args)
|
||||
@@ -0,0 +1,35 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
from diffsynth import load_state_dict
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
state_dict = load_state_dict("models/train/Ace-Step1.5_full/epoch-1.safetensors")
|
||||
pipe.dit.load_state_dict(state_dict)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "Ace-Step1.5_full.wav")
|
||||
@@ -0,0 +1,35 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
from diffsynth import load_state_dict
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-base", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
state_dict = load_state_dict("models/train/acestep-v15-base_full/epoch-1.safetensors")
|
||||
pipe.dit.load_state_dict(state_dict)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=42,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-base_full.wav")
|
||||
@@ -0,0 +1,35 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
from diffsynth import load_state_dict
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-sft", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
state_dict = load_state_dict("models/train/acestep-v15-sft_full/epoch-1.safetensors")
|
||||
pipe.dit.load_state_dict(state_dict)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-sft_full.wav")
|
||||
@@ -0,0 +1,35 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
from diffsynth import load_state_dict
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-turbo-continuous", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
state_dict = load_state_dict("models/train/acestep-v15-turbo-continuous_full/epoch-1.safetensors")
|
||||
pipe.dit.load_state_dict(state_dict)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-turbo-continuous_full.wav")
|
||||
@@ -0,0 +1,35 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
from diffsynth import load_state_dict
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-turbo-shift1", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
state_dict = load_state_dict("models/train/acestep-v15-turbo-shift1_full/epoch-1.safetensors")
|
||||
pipe.dit.load_state_dict(state_dict)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-turbo-shift1_full.wav")
|
||||
@@ -0,0 +1,35 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
from diffsynth import load_state_dict
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-turbo-shift3", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
state_dict = load_state_dict("models/train/acestep-v15-turbo-shift3_full/epoch-1.safetensors")
|
||||
pipe.dit.load_state_dict(state_dict)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-turbo-shift3_full.wav")
|
||||
@@ -0,0 +1,35 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
from diffsynth import load_state_dict
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-xl-base", origin_file_pattern="model-*.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
state_dict = load_state_dict("models/train/acestep-v15-xl-base_full/epoch-1.safetensors")
|
||||
pipe.dit.load_state_dict(state_dict)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-xl-base_full.wav")
|
||||
@@ -0,0 +1,35 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
from diffsynth import load_state_dict
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-xl-sft", origin_file_pattern="model-*.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
state_dict = load_state_dict("models/train/acestep-v15-xl-sft_full/epoch-1.safetensors")
|
||||
pipe.dit.load_state_dict(state_dict)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-xl-sft_full.wav")
|
||||
@@ -0,0 +1,33 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
pipe.load_lora(pipe.dit, "models/train/Ace-Step1.5_lora/epoch-9.safetensors", alpha=1)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "Ace-Step1.5_lora.wav")
|
||||
@@ -0,0 +1,33 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-base", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
pipe.load_lora(pipe.dit, "models/train/acestep-v15-base_lora/epoch-9.safetensors", alpha=1)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-base_lora.wav")
|
||||
@@ -0,0 +1,33 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-sft", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
pipe.load_lora(pipe.dit, "models/train/acestep-v15-sft_lora/epoch-9.safetensors", alpha=1)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-sft_lora.wav")
|
||||
@@ -0,0 +1,33 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-turbo-continuous", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
pipe.load_lora(pipe.dit, "models/train/acestep-v15-turbo-continuous_lora/epoch-9.safetensors", alpha=1)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-turbo-continuous_lora.wav")
|
||||
@@ -0,0 +1,33 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-turbo-shift1", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
pipe.load_lora(pipe.dit, "models/train/acestep-v15-turbo-shift1_lora/epoch-9.safetensors", alpha=1)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-turbo-shift1_lora.wav")
|
||||
@@ -0,0 +1,33 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-turbo-shift3", origin_file_pattern="model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
pipe.load_lora(pipe.dit, "models/train/acestep-v15-turbo-shift3_lora/epoch-9.safetensors", alpha=1)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-turbo-shift3_lora.wav")
|
||||
@@ -0,0 +1,33 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-xl-base", origin_file_pattern="model-*.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
pipe.load_lora(pipe.dit, "models/train/acestep-v15-xl-base_lora/epoch-9.safetensors", alpha=1)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-xl-base_lora.wav")
|
||||
@@ -0,0 +1,33 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-xl-sft", origin_file_pattern="model-*.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
pipe.load_lora(pipe.dit, "models/train/acestep-v15-xl-sft_lora/epoch-9.safetensors", alpha=1)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-xl-sft_lora.wav")
|
||||
@@ -0,0 +1,33 @@
|
||||
from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
|
||||
from diffsynth.utils.data.audio import save_audio
|
||||
import torch
|
||||
|
||||
|
||||
pipe = AceStepPipeline.from_pretrained(
|
||||
torch_dtype=torch.bfloat16,
|
||||
device="cuda",
|
||||
model_configs=[
|
||||
ModelConfig(model_id="ACE-Step/acestep-v15-xl-turbo", origin_file_pattern="model-*.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
|
||||
ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
|
||||
],
|
||||
text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
|
||||
silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
|
||||
)
|
||||
pipe.load_lora(pipe.dit, "models/train/acestep-v15-xl-turbo_lora/epoch-9.safetensors", alpha=1)
|
||||
|
||||
prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel. The song kicks off with a catchy, synthesized brass fanfare over a driving rock beat with punchy drums and a solid bassline. A powerful, clear male vocal enters with a theatrical and energetic delivery, soaring through the verses and hitting powerful high notes in the chorus. The arrangement is dense and dynamic, featuring rhythmic electric guitar chords, brief instrumental breaks with synth flourishes, and a consistent, danceable groove throughout. The overall mood is triumphant, adventurous, and exhilarating."
|
||||
lyrics = '[Intro - Synth Brass Fanfare]\n\n[Verse 1]\n黑夜里的风吹过耳畔\n甜蜜时光转瞬即万\n脚步飘摇在星光上\n心追节奏心跳狂乱\n耳边传来电吉他呼唤\n手指轻触碰点流点燃\n梦在云端任它蔓延\n疯狂跳跃自由无间\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Instrumental Break - Synth Brass Melody]\n\n[Verse 2]\n鼓点撞击黑夜的底端\n跳动节拍连接你我俩\n在这里让灵魂发光\n燃尽所有不留遗憾\n\n[Instrumental Break - Synth Brass Melody]\n\n[Bridge]\n光影交错彼此的视线\n霓虹之下夜空的蔚蓝\n月光洒下温热心田\n追逐梦想它不会遥远\n\n[Chorus]\n心电感应在震动间\n拥抱未来勇敢冒险\n那旋律在心中无限\n世界变得如此耀眼\n\n[Outro - Instrumental with Synth Brass Melody]\n[Song ends abruptly]'
|
||||
audio = pipe(
|
||||
prompt=prompt,
|
||||
lyrics=lyrics,
|
||||
duration=160,
|
||||
bpm=100,
|
||||
keyscale="B minor",
|
||||
timesignature="4",
|
||||
vocal_language="zh",
|
||||
seed=1,
|
||||
num_inference_steps=30,
|
||||
cfg_scale=4.0,
|
||||
)
|
||||
save_audio(audio, pipe.vae.sampling_rate, "acestep-v15-xl-turbo_lora.wav")
|
||||
Reference in New Issue
Block a user